diff --git a/changelog/3918.added.md b/changelog/3918.added.md
new file mode 100644
index 000000000..6b4d4446f
--- /dev/null
+++ b/changelog/3918.added.md
@@ -0,0 +1,15 @@
+- Wired up `system_instruction` in `BaseOpenAILLMService`, `AnthropicLLMService`, and `AWSBedrockLLMService` so it works as a default system prompt, matching the behavior of the Google services. This enables sharing a single `LLMContext` across multiple LLM services, where each service provides its own system instruction independently.
+
+  ```python
+  llm = OpenAILLMService(
+      api_key=os.getenv("OPENAI_API_KEY"),
+      system_instruction="You are a helpful assistant.",
+  )
+
+  context = LLMContext()
+
+  @transport.event_handler("on_client_connected")
+  async def on_client_connected(transport, client):
+      context.add_message({"role": "user", "content": "Please introduce yourself."})
+      await task.queue_frames([LLMRunFrame()])
+  ```
diff --git a/changelog/3918.other.md b/changelog/3918.other.md
new file mode 100644
index 000000000..6caa1ba05
--- /dev/null
+++ b/changelog/3918.other.md
@@ -0,0 +1 @@
+- Updated foundational examples to use `system_instruction` on LLM services instead of adding system messages to `LLMContext`.
diff --git a/examples/foundational/02-llm-say-one-thing.py b/examples/foundational/02-llm-say-one-thing.py
index f4fbfcbe0..9cfd7c39e 100644
--- a/examples/foundational/02-llm-say-one-thing.py
+++ b/examples/foundational/02-llm-say-one-thing.py
@@ -42,14 +42,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
-
-    messages = [
-        {
-            "role": "system",
-            "content": "You are an LLM in a WebRTC session, and this is a 'hello world' demo. Say hello to the world.",
-        }
-    ]
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are an LLM in a WebRTC session, and this is a 'hello world' demo.",
+    )
 
     task = PipelineTask(
         Pipeline([llm, tts, transport.output()]),
@@ -59,7 +55,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     # Register an event handler so we can play the audio when the client joins
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
-        await task.queue_frames([LLMContextFrame(LLMContext(messages)), EndFrame()])
+        context = LLMContext()
+        context.add_message({"role": "system", "content": "Say hello to the world."})
+        await task.queue_frames([LLMContextFrame(context), EndFrame()])
 
     runner = PipelineRunner(handle_sigint=runner_args.handle_sigint)
 
diff --git a/examples/foundational/04-transports-small-webrtc.py b/examples/foundational/04-transports-small-webrtc.py
index 7196dbb05..d29171dd2 100644
--- a/examples/foundational/04-transports-small-webrtc.py
+++ b/examples/foundational/04-transports-small-webrtc.py
@@ -70,16 +70,12 @@ async def run_example(webrtc_connection: SmallWebRTCConnection):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -109,7 +105,7 @@ async def run_example(webrtc_connection: SmallWebRTCConnection):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/04a-transports-daily.py b/examples/foundational/04a-transports-daily.py
index d1d88dbec..b0a7958ef 100644
--- a/examples/foundational/04a-transports-daily.py
+++ b/examples/foundational/04a-transports-daily.py
@@ -53,16 +53,13 @@ async def main():
             voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4o")
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            model="gpt-4o",
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -91,7 +88,9 @@ async def main():
         async def on_first_participant_joined(transport, participant):
             await transport.capture_participant_transcription(participant["id"])
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_participant_left")
diff --git a/examples/foundational/04b-transports-livekit.py b/examples/foundational/04b-transports-livekit.py
index b577b453e..e05c5ddc3 100644
--- a/examples/foundational/04b-transports-livekit.py
+++ b/examples/foundational/04b-transports-livekit.py
@@ -55,24 +55,17 @@ async def main():
 
     stt = DeepgramSTTService(api_key=os.getenv("DEEPGRAM_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     tts = CartesiaTTSService(
         api_key=os.getenv("CARTESIA_API_KEY"),
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. "
-            "Your goal is to demonstrate your capabilities in a succinct way. "
-            "Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. "
-            "Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/06-listen-and-respond.py b/examples/foundational/06-listen-and-respond.py
index 14dee63ad..418b3b5af 100644
--- a/examples/foundational/06-listen-and-respond.py
+++ b/examples/foundational/06-listen-and-respond.py
@@ -86,18 +86,14 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     ml = MetricsLogger()
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -129,7 +125,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/06a-image-sync.py b/examples/foundational/06a-image-sync.py
index be0e4fe78..f963aa339 100644
--- a/examples/foundational/06a-image-sync.py
+++ b/examples/foundational/06a-image-sync.py
@@ -103,16 +103,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/07-interruptible-cartesia-http.py b/examples/foundational/07-interruptible-cartesia-http.py
index e65fb3750..558217413 100644
--- a/examples/foundational/07-interruptible-cartesia-http.py
+++ b/examples/foundational/07-interruptible-cartesia-http.py
@@ -59,16 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +95,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07-interruptible.py b/examples/foundational/07-interruptible.py
index 074e091ea..3f97e5028 100644
--- a/examples/foundational/07-interruptible.py
+++ b/examples/foundational/07-interruptible.py
@@ -62,16 +62,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         # text_aggregation_mode=TextAggregationMode.TOKEN,
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +98,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07c-interruptible-deepgram-flux.py b/examples/foundational/07c-interruptible-deepgram-flux.py
index d2bcceaf7..3f2d6e28d 100644
--- a/examples/foundational/07c-interruptible-deepgram-flux.py
+++ b/examples/foundational/07c-interruptible-deepgram-flux.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = DeepgramTTSService(api_key=os.getenv("DEEPGRAM_API_KEY"), voice="aura-2-andromeda-en")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(
@@ -104,7 +100,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07c-interruptible-deepgram-http.py b/examples/foundational/07c-interruptible-deepgram-http.py
index 53fc77722..6955ccbd1 100644
--- a/examples/foundational/07c-interruptible-deepgram-http.py
+++ b/examples/foundational/07c-interruptible-deepgram-http.py
@@ -63,14 +63,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
+        messages = []
 
         context = LLMContext(messages)
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
@@ -103,7 +101,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07c-interruptible-deepgram-sagemaker.py b/examples/foundational/07c-interruptible-deepgram-sagemaker.py
index 7a4cd4297..8c7fc6556 100644
--- a/examples/foundational/07c-interruptible-deepgram-sagemaker.py
+++ b/examples/foundational/07c-interruptible-deepgram-sagemaker.py
@@ -76,16 +76,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         aws_region=os.getenv("AWS_REGION"),
         model="us.amazon.nova-pro-v1:0",
         params=AWSBedrockLLMService.InputParams(temperature=0.8),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -116,7 +110,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07c-interruptible-deepgram-vad.py b/examples/foundational/07c-interruptible-deepgram-vad.py
index 62fb29dff..aaa81fc90 100644
--- a/examples/foundational/07c-interruptible-deepgram-vad.py
+++ b/examples/foundational/07c-interruptible-deepgram-vad.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = DeepgramTTSService(api_key=os.getenv("DEEPGRAM_API_KEY"), voice="aura-2-andromeda-en")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(user_turn_strategies=ExternalUserTurnStrategies()),
@@ -101,7 +97,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07c-interruptible-deepgram.py b/examples/foundational/07c-interruptible-deepgram.py
index 0133ad150..1cbfc72cc 100644
--- a/examples/foundational/07c-interruptible-deepgram.py
+++ b/examples/foundational/07c-interruptible-deepgram.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = DeepgramTTSService(api_key=os.getenv("DEEPGRAM_API_KEY"), voice="aura-2-andromeda-en")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07d-interruptible-elevenlabs-http.py b/examples/foundational/07d-interruptible-elevenlabs-http.py
index 0299c1300..85cffc270 100644
--- a/examples/foundational/07d-interruptible-elevenlabs-http.py
+++ b/examples/foundational/07d-interruptible-elevenlabs-http.py
@@ -67,16 +67,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -107,7 +103,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07d-interruptible-elevenlabs.py b/examples/foundational/07d-interruptible-elevenlabs.py
index feaf41255..b55345f33 100644
--- a/examples/foundational/07d-interruptible-elevenlabs.py
+++ b/examples/foundational/07d-interruptible-elevenlabs.py
@@ -60,16 +60,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id=os.getenv("ELEVENLABS_VOICE_ID", ""),
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07f-interruptible-azure-http.py b/examples/foundational/07f-interruptible-azure-http.py
index 3d8bdf073..1946f66ee 100644
--- a/examples/foundational/07f-interruptible-azure-http.py
+++ b/examples/foundational/07f-interruptible-azure-http.py
@@ -66,16 +66,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         api_key=os.getenv("AZURE_CHATGPT_API_KEY"),
         endpoint=os.getenv("AZURE_CHATGPT_ENDPOINT"),
         model=os.getenv("AZURE_CHATGPT_MODEL"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -106,7 +100,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07f-interruptible-azure.py b/examples/foundational/07f-interruptible-azure.py
index 03b399a5a..68f5a72f9 100644
--- a/examples/foundational/07f-interruptible-azure.py
+++ b/examples/foundational/07f-interruptible-azure.py
@@ -66,16 +66,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         api_key=os.getenv("AZURE_CHATGPT_API_KEY"),
         endpoint=os.getenv("AZURE_CHATGPT_ENDPOINT"),
         model=os.getenv("AZURE_CHATGPT_MODEL"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -106,7 +100,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07g-interruptible-openai-http.py b/examples/foundational/07g-interruptible-openai-http.py
index 65b2f8b9b..05c98ca59 100644
--- a/examples/foundational/07g-interruptible-openai-http.py
+++ b/examples/foundational/07g-interruptible-openai-http.py
@@ -60,16 +60,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = OpenAITTSService(api_key=os.getenv("OPENAI_API_KEY"), voice="ballad")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are very knowledgable about dogs. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are very knowledgable about dogs. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -101,7 +97,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07g-interruptible-openai.py b/examples/foundational/07g-interruptible-openai.py
index 551742a57..a29e8210f 100644
--- a/examples/foundational/07g-interruptible-openai.py
+++ b/examples/foundational/07g-interruptible-openai.py
@@ -66,16 +66,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = OpenAITTSService(api_key=os.getenv("OPENAI_API_KEY"), voice="ballad")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are very knowledgable about dogs. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are very knowledgable about dogs. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -107,7 +103,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07h-interruptible-openpipe.py b/examples/foundational/07h-interruptible-openpipe.py
index 052ddfd14..e3ddc05b8 100644
--- a/examples/foundational/07h-interruptible-openpipe.py
+++ b/examples/foundational/07h-interruptible-openpipe.py
@@ -65,16 +65,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         api_key=os.getenv("OPENAI_API_KEY"),
         openpipe_api_key=os.getenv("OPENPIPE_API_KEY"),
         tags={"conversation_id": f"pipecat-{timestamp}"},
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -105,7 +99,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07i-interruptible-xtts.py b/examples/foundational/07i-interruptible-xtts.py
index 8b2b3d6d5..00883e37f 100644
--- a/examples/foundational/07i-interruptible-xtts.py
+++ b/examples/foundational/07i-interruptible-xtts.py
@@ -63,16 +63,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             base_url="http://localhost:8000",
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -103,7 +99,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07k-interruptible-lmnt.py b/examples/foundational/07k-interruptible-lmnt.py
index a6d1c56b4..246caee19 100644
--- a/examples/foundational/07k-interruptible-lmnt.py
+++ b/examples/foundational/07k-interruptible-lmnt.py
@@ -56,16 +56,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = LmntTTSService(api_key=os.getenv("LMNT_API_KEY"), voice_id="morgan")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07l-interruptible-groq.py b/examples/foundational/07l-interruptible-groq.py
index 5c92814e3..5a16d789e 100644
--- a/examples/foundational/07l-interruptible-groq.py
+++ b/examples/foundational/07l-interruptible-groq.py
@@ -55,19 +55,14 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     stt = GroqSTTService(api_key=os.getenv("GROQ_API_KEY"))
 
     llm = GroqLLMService(
-        api_key=os.getenv("GROQ_API_KEY"), model="meta-llama/llama-4-maverick-17b-128e-instruct"
+        api_key=os.getenv("GROQ_API_KEY"),
+        model="meta-llama/llama-4-maverick-17b-128e-instruct",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
     tts = GroqTTSService(api_key=os.getenv("GROQ_API_KEY"))
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -98,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07m-interruptible-aws.py b/examples/foundational/07m-interruptible-aws.py
index eac193531..b30220bcd 100644
--- a/examples/foundational/07m-interruptible-aws.py
+++ b/examples/foundational/07m-interruptible-aws.py
@@ -62,16 +62,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         aws_region="us-west-2",
         model="us.anthropic.claude-haiku-4-5-20251001-v1:0",
         params=AWSBedrockLLMService.InputParams(temperature=0.8),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "user", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07n-interruptible-gemini-image.py b/examples/foundational/07n-interruptible-gemini-image.py
index 21fea9d1b..b7d3281b9 100644
--- a/examples/foundational/07n-interruptible-gemini-image.py
+++ b/examples/foundational/07n-interruptible-gemini-image.py
@@ -83,17 +83,11 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     llm = GoogleLLMService(
         api_key=os.getenv("GOOGLE_API_KEY"),
         model="gemini-2.5-flash-image",
-        # model="gemini-3-pro-image-preview", # A more powerful model, but slower
+        # model="gemini-3-pro-image-preview", # A more powerful model, but slower,
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -124,7 +118,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation with a styled introduction
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07n-interruptible-gemini.py b/examples/foundational/07n-interruptible-gemini.py
index 08315fcfa..74f732926 100644
--- a/examples/foundational/07n-interruptible-gemini.py
+++ b/examples/foundational/07n-interruptible-gemini.py
@@ -71,13 +71,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     llm = GoogleLLMService(
         api_key=os.getenv("GOOGLE_API_KEY"),
         model="gemini-2.5-flash",
-    )
-
-    # System message that instructs the AI on how to speak
-    messages = [
-        {
-            "role": "system",
-            "content": """You are a helpful AI assistant in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way.
+        system_instruction="""You are a helpful AI assistant in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way.
 
             IMPORTANT: You're using Gemini TTS which supports expressive markup tags. You can use these tags in your responses:
             - [sigh] - Insert a sigh sound
@@ -95,10 +89,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             - "The answer is... [long pause] ...42!"
 
             Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.""",
-        },
-    ]
+    )
 
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -129,7 +122,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation
-        messages.append(
+        context.add_message(
             {
                 "role": "system",
                 "content": "You are an AI assistant. You can help with a variety of tasks. Introduce yourself and ask the user what they would like to know.",
diff --git a/examples/foundational/07n-interruptible-google-http.py b/examples/foundational/07n-interruptible-google-http.py
index e1fd0002f..e5f82ffc5 100644
--- a/examples/foundational/07n-interruptible-google-http.py
+++ b/examples/foundational/07n-interruptible-google-http.py
@@ -72,16 +72,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         # params=GoogleLLMService.InputParams(
         #     thinking=GoogleLLMService.ThinkingConfig(thinking_budget=4096)
         # ),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -112,7 +106,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07n-interruptible-google.py b/examples/foundational/07n-interruptible-google.py
index e5c792c12..090fdbbdc 100644
--- a/examples/foundational/07n-interruptible-google.py
+++ b/examples/foundational/07n-interruptible-google.py
@@ -72,16 +72,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         # params=GoogleLLMService.InputParams(
         #     thinking=GoogleLLMService.ThinkingConfig(thinking_budget=4096)
         # ),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -112,7 +106,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07o-interruptible-assemblyai-turn-detection.py b/examples/foundational/07o-interruptible-assemblyai-turn-detection.py
index cf052ada4..13f596136 100644
--- a/examples/foundational/07o-interruptible-assemblyai-turn-detection.py
+++ b/examples/foundational/07o-interruptible-assemblyai-turn-detection.py
@@ -111,16 +111,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(
@@ -154,7 +150,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07o-interruptible-assemblyai.py b/examples/foundational/07o-interruptible-assemblyai.py
index b65116b46..8378f7bba 100644
--- a/examples/foundational/07o-interruptible-assemblyai.py
+++ b/examples/foundational/07o-interruptible-assemblyai.py
@@ -62,16 +62,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +98,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07p-interruptible-krisp-viva.py b/examples/foundational/07p-interruptible-krisp-viva.py
index 24929a825..4df652ae0 100644
--- a/examples/foundational/07p-interruptible-krisp-viva.py
+++ b/examples/foundational/07p-interruptible-krisp-viva.py
@@ -87,16 +87,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         api_key=os.getenv("CARTESIA_API_KEY"), voice_id="71a7ad14-091c-4e8e-a314-022ece01c121"
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(
@@ -133,7 +129,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07p-interruptible-krisp.py b/examples/foundational/07p-interruptible-krisp.py
index 6a17a57eb..ad737ba05 100644
--- a/examples/foundational/07p-interruptible-krisp.py
+++ b/examples/foundational/07p-interruptible-krisp.py
@@ -60,16 +60,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = DeepgramTTSService(api_key=os.getenv("DEEPGRAM_API_KEY"), voice="aura-helios-en")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07q-interruptible-rime-http.py b/examples/foundational/07q-interruptible-rime-http.py
index 902b6b231..4e390b123 100644
--- a/examples/foundational/07q-interruptible-rime-http.py
+++ b/examples/foundational/07q-interruptible-rime-http.py
@@ -65,16 +65,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -105,7 +101,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07q-interruptible-rime.py b/examples/foundational/07q-interruptible-rime.py
index 27e010273..cebf6bfd0 100644
--- a/examples/foundational/07q-interruptible-rime.py
+++ b/examples/foundational/07q-interruptible-rime.py
@@ -59,16 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="luna",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +95,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07r-interruptible-nvidia.py b/examples/foundational/07r-interruptible-nvidia.py
index d3e34c61f..f50667267 100644
--- a/examples/foundational/07r-interruptible-nvidia.py
+++ b/examples/foundational/07r-interruptible-nvidia.py
@@ -57,18 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     llm = NvidiaLLMService(
         api_key=os.getenv("NVIDIA_API_KEY"),
         model="meta/llama-3.3-70b-instruct",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
     tts = NvidiaTTSService(api_key=os.getenv("NVIDIA_API_KEY"))
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07t-interruptible-fish.py b/examples/foundational/07t-interruptible-fish.py
index 2df978475..8b4152103 100644
--- a/examples/foundational/07t-interruptible-fish.py
+++ b/examples/foundational/07t-interruptible-fish.py
@@ -60,16 +60,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         model="4ce7e917cedd4bc2bb2e6ff3a46acaa1",  # Barack Obama
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07v-interruptible-neuphonic-http.py b/examples/foundational/07v-interruptible-neuphonic-http.py
index d3f27613e..60a08e623 100644
--- a/examples/foundational/07v-interruptible-neuphonic-http.py
+++ b/examples/foundational/07v-interruptible-neuphonic-http.py
@@ -64,16 +64,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -104,7 +100,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07v-interruptible-neuphonic.py b/examples/foundational/07v-interruptible-neuphonic.py
index c1ee64f3e..02d6862a8 100644
--- a/examples/foundational/07v-interruptible-neuphonic.py
+++ b/examples/foundational/07v-interruptible-neuphonic.py
@@ -59,16 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="fc854436-2dac-4d21-aa69-ae17b54e98eb",  # Emily
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +95,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07w-interruptible-fal.py b/examples/foundational/07w-interruptible-fal.py
index 41fad024a..2d1886ddb 100644
--- a/examples/foundational/07w-interruptible-fal.py
+++ b/examples/foundational/07w-interruptible-fal.py
@@ -62,16 +62,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +98,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07x-interruptible-local.py b/examples/foundational/07x-interruptible-local.py
index 65e5836bc..3465f67b2 100644
--- a/examples/foundational/07x-interruptible-local.py
+++ b/examples/foundational/07x-interruptible-local.py
@@ -47,16 +47,12 @@ async def main():
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -82,7 +78,7 @@ async def main():
         ),
     )
 
-    messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+    context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
     await task.queue_frames([LLMRunFrame()])
 
     runner = PipelineRunner()
diff --git a/examples/foundational/07y-interruptible-minimax.py b/examples/foundational/07y-interruptible-minimax.py
index 1900b1ba8..f23d86f34 100644
--- a/examples/foundational/07y-interruptible-minimax.py
+++ b/examples/foundational/07y-interruptible-minimax.py
@@ -66,16 +66,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             params=MiniMaxHttpTTSService.InputParams(language=Language.EN),
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -106,7 +102,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07z-interruptible-sarvam-http.py b/examples/foundational/07z-interruptible-sarvam-http.py
index e65be1f8e..1f9a8a4e3 100644
--- a/examples/foundational/07z-interruptible-sarvam-http.py
+++ b/examples/foundational/07z-interruptible-sarvam-http.py
@@ -68,16 +68,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             params=SarvamHttpTTSService.InputParams(language=Language.EN),
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -108,7 +104,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07z-interruptible-sarvam.py b/examples/foundational/07z-interruptible-sarvam.py
index 82e070be2..a9a03e7d4 100644
--- a/examples/foundational/07z-interruptible-sarvam.py
+++ b/examples/foundational/07z-interruptible-sarvam.py
@@ -62,16 +62,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         model="bulbul:v2",
         voice_id="manisha",
     )
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -101,7 +97,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         # Optionally, you can wait for 30 seconds and then change the voice.
diff --git a/examples/foundational/07za-interruptible-soniox.py b/examples/foundational/07za-interruptible-soniox.py
index a60cfe992..f2d651e5b 100644
--- a/examples/foundational/07za-interruptible-soniox.py
+++ b/examples/foundational/07za-interruptible-soniox.py
@@ -64,16 +64,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -103,7 +99,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zb-interruptible-inworld-http.py b/examples/foundational/07zb-interruptible-inworld-http.py
index c6441e1c3..9ef465781 100644
--- a/examples/foundational/07zb-interruptible-inworld-http.py
+++ b/examples/foundational/07zb-interruptible-inworld-http.py
@@ -64,16 +64,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             streaming=True,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful AI demonstrating Inworld AI's TTS. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a friendly and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful AI demonstrating Inworld AI's TTS. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a friendly and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -111,7 +107,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info("Client connected")
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zb-interruptible-inworld.py b/examples/foundational/07zb-interruptible-inworld.py
index 9af1302d3..7cdbfd88a 100644
--- a/examples/foundational/07zb-interruptible-inworld.py
+++ b/examples/foundational/07zb-interruptible-inworld.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         temperature=1.1,
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful AI demonstrating Inworld AI's TTS. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a friendly and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful AI demonstrating Inworld AI's TTS. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a friendly and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -108,7 +104,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info("Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zc-interruptible-asyncai-http.py b/examples/foundational/07zc-interruptible-asyncai-http.py
index 3720736bd..17d187ac8 100644
--- a/examples/foundational/07zc-interruptible-asyncai-http.py
+++ b/examples/foundational/07zc-interruptible-asyncai-http.py
@@ -64,16 +64,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -104,7 +100,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
         @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zc-interruptible-asyncai.py b/examples/foundational/07zc-interruptible-asyncai.py
index 31d8fb437..799842658 100644
--- a/examples/foundational/07zc-interruptible-asyncai.py
+++ b/examples/foundational/07zc-interruptible-asyncai.py
@@ -60,16 +60,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id=os.getenv("ASYNCAI_VOICE_ID", "e0f39dc4-f691-4e78-bba5-5c636692cc04"),
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zd-interruptible-aicoustics.py b/examples/foundational/07zd-interruptible-aicoustics.py
index 1a3c626ef..9e9f6ddd6 100644
--- a/examples/foundational/07zd-interruptible-aicoustics.py
+++ b/examples/foundational/07zd-interruptible-aicoustics.py
@@ -80,16 +80,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=aic_vad_analyzer),
@@ -128,7 +124,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         logger.info(f"Client connected")
         await audiobuffer.start_recording()
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @audiobuffer.event_handler("on_audio_data")
diff --git a/examples/foundational/07ze-interruptible-hume.py b/examples/foundational/07ze-interruptible-hume.py
index 820e12a73..395aa75d4 100644
--- a/examples/foundational/07ze-interruptible-hume.py
+++ b/examples/foundational/07ze-interruptible-hume.py
@@ -62,16 +62,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="f898a92e-685f-43fa-985b-a46920f0650b",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -113,7 +109,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             "💡 Word timestamps are enabled! Watch the console for TTSTextFrame logs showing each word with its PTS."
         )
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zf-interruptible-gradium.py b/examples/foundational/07zf-interruptible-gradium.py
index 7c823325f..d76735897 100644
--- a/examples/foundational/07zf-interruptible-gradium.py
+++ b/examples/foundational/07zf-interruptible-gradium.py
@@ -66,16 +66,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         url="wss://us.api.gradium.ai/api/speech/tts",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -106,7 +102,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zg-interruptible-camb.py b/examples/foundational/07zg-interruptible-camb.py
index ace1fd8b3..669315603 100644
--- a/examples/foundational/07zg-interruptible-camb.py
+++ b/examples/foundational/07zg-interruptible-camb.py
@@ -59,18 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         model="mars-flash",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful voice assistant powered by Camb AI text-to-speech. ",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful voice assistant powered by Camb AI text-to-speech. "
-            "Keep your responses concise and conversational since they will be spoken aloud. "
-            "Avoid special characters, emojis, or bullet points.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -101,7 +95,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info("Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zh-interruptible-hathora.py b/examples/foundational/07zh-interruptible-hathora.py
index 7e903035e..08749bf1a 100644
--- a/examples/foundational/07zh-interruptible-hathora.py
+++ b/examples/foundational/07zh-interruptible-hathora.py
@@ -64,16 +64,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         base_url="https://app-362f7ca1-6975-4e18-a605-ab202bf2c315.app.hathora.dev/v1",
         api_key=os.getenv("HATHORA_API_KEY"),
         model=None,
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     context_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -104,7 +98,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zi-interruptible-piper.py b/examples/foundational/07zi-interruptible-piper.py
index 73b8d0f84..9d07f9cdf 100644
--- a/examples/foundational/07zi-interruptible-piper.py
+++ b/examples/foundational/07zi-interruptible-piper.py
@@ -56,16 +56,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = PiperTTSService(voice_id="en_US-ryan-high")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zj-interruptible-kokoro.py b/examples/foundational/07zj-interruptible-kokoro.py
index 685388813..04ab90145 100644
--- a/examples/foundational/07zj-interruptible-kokoro.py
+++ b/examples/foundational/07zj-interruptible-kokoro.py
@@ -56,16 +56,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = KokoroTTSService(voice_id="af_heart")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/07zk-interruptible-resemble.py b/examples/foundational/07zk-interruptible-resemble.py
index 3594d5b4f..d9298815d 100644
--- a/examples/foundational/07zk-interruptible-resemble.py
+++ b/examples/foundational/07zk-interruptible-resemble.py
@@ -62,16 +62,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id=os.getenv("RESEMBLE_VOICE_UUID"),
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +98,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/08-custom-frame-processor.py b/examples/foundational/08-custom-frame-processor.py
index ba55eec7d..a4b5b2127 100644
--- a/examples/foundational/08-custom-frame-processor.py
+++ b/examples/foundational/08-custom-frame-processor.py
@@ -98,16 +98,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -141,7 +137,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected: {client}")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/10-wake-phrase.py b/examples/foundational/10-wake-phrase.py
index c7f53e496..52a7d1225 100644
--- a/examples/foundational/10-wake-phrase.py
+++ b/examples/foundational/10-wake-phrase.py
@@ -10,7 +10,7 @@ from dotenv import load_dotenv
 from loguru import logger
 
 from pipecat.audio.vad.silero import SileroVADAnalyzer
-from pipecat.frames.frames import TTSSpeakFrame
+from pipecat.frames.frames import LLMRunFrame, TTSSpeakFrame
 from pipecat.pipeline.pipeline import Pipeline
 from pipecat.pipeline.runner import PipelineRunner
 from pipecat.pipeline.task import PipelineParams, PipelineTask
@@ -59,18 +59,14 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
-
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful assistant. Respond to what the user said in a creative and helpful way. Keep your responses brief.",
-        },
-    ]
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful assistant. Respond to what the user said in a creative and helpful way. Keep your responses brief.",
+    )
 
     hey_robot_filter = WakeCheckFilter(["hey robot", "hey, robot"])
 
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +98,13 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        await task.queue_frame(TTSSpeakFrame("Hi! If you want to talk to me, just say 'Hey Robot'"))
+        context.add_message(
+            {
+                "role": "system",
+                "content": "Please introduce yourself. Tell the user they should say 'Hey Robot' before talking to you.",
+            }
+        )
+        await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
     async def on_client_disconnected(transport, client):
diff --git a/examples/foundational/11-sound-effects.py b/examples/foundational/11-sound-effects.py
index 156778317..401c36b8e 100644
--- a/examples/foundational/11-sound-effects.py
+++ b/examples/foundational/11-sound-effects.py
@@ -104,21 +104,17 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     stt = DeepgramSTTService(api_key=os.getenv("DEEPGRAM_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     tts = CartesiaTTSService(
         api_key=os.getenv("CARTESIA_API_KEY"),
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/12-describe-image-openai.py b/examples/foundational/12-describe-image-openai.py
index afa334429..dc45451fe 100644
--- a/examples/foundational/12-describe-image-openai.py
+++ b/examples/foundational/12-describe-image-openai.py
@@ -56,16 +56,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are also able to describe images.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are also able to describe images.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -114,7 +110,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             size=image.size,
             text=question,
         )
-        messages.append(message)
+        context.add_message(message)
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/12a-describe-image-anthropic.py b/examples/foundational/12a-describe-image-anthropic.py
index 2b34c6220..a1151ddf9 100644
--- a/examples/foundational/12a-describe-image-anthropic.py
+++ b/examples/foundational/12a-describe-image-anthropic.py
@@ -56,16 +56,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = AnthropicLLMService(api_key=os.getenv("ANTHROPIC_API_KEY"))
+    llm = AnthropicLLMService(
+        api_key=os.getenv("ANTHROPIC_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are also able to describe images.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are also able to describe images.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -114,7 +110,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             size=image.size,
             text=question,
         )
-        messages.append(message)
+        context.add_message(message)
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/12b-describe-image-aws.py b/examples/foundational/12b-describe-image-aws.py
index 84a391121..43a6ae232 100644
--- a/examples/foundational/12b-describe-image-aws.py
+++ b/examples/foundational/12b-describe-image-aws.py
@@ -63,16 +63,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         # Here we can't because AWS Bedrock doesn't support it for Claude 3.7,
         # which we need for image input.
         params=AWSBedrockLLMService.InputParams(temperature=0.8),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are also able to describe images.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are also able to describe images.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -121,7 +115,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             size=image.size,
             text=question,
         )
-        messages.append(message)
+        context.add_message(message)
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/12c-describe-image-gemini-flash.py b/examples/foundational/12c-describe-image-gemini-flash.py
index 212dc5b32..9adbe62d2 100644
--- a/examples/foundational/12c-describe-image-gemini-flash.py
+++ b/examples/foundational/12c-describe-image-gemini-flash.py
@@ -56,16 +56,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = GoogleLLMService(api_key=os.getenv("GOOGLE_API_KEY"))
+    llm = GoogleLLMService(
+        api_key=os.getenv("GOOGLE_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are also able to describe images.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are also able to describe images.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -114,7 +110,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             size=image.size,
             text=question,
         )
-        messages.append(message)
+        context.add_message(message)
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/14-function-calling.py b/examples/foundational/14-function-calling.py
index 27e666d0b..0087af9fc 100644
--- a/examples/foundational/14-function-calling.py
+++ b/examples/foundational/14-function-calling.py
@@ -70,7 +70,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -110,14 +113,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function, restaurant_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14c-function-calling-together.py b/examples/foundational/14c-function-calling-together.py
index 28dd8c466..214d6f292 100644
--- a/examples/foundational/14c-function-calling-together.py
+++ b/examples/foundational/14c-function-calling-together.py
@@ -70,6 +70,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     llm = TogetherLLMService(
         api_key=os.getenv("TOGETHER_API_KEY"),
         model="meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -96,14 +97,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["location", "format"],
     )
     tools = ToolsSchema(standard_tools=[weather_function])
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14d-function-calling-anthropic-video.py b/examples/foundational/14d-function-calling-anthropic-video.py
index 630a0aaac..cddde3647 100644
--- a/examples/foundational/14d-function-calling-anthropic-video.py
+++ b/examples/foundational/14d-function-calling-anthropic-video.py
@@ -94,7 +94,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
 
     # Anthropic for vision analysis
-    llm = AnthropicLLMService(api_key=os.getenv("ANTHROPIC_API_KEY"))
+    llm = AnthropicLLMService(
+        api_key=os.getenv("ANTHROPIC_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
+    )
     llm.register_function("fetch_user_image", fetch_user_image)
 
     @llm.event_handler("on_function_calls_started")
@@ -118,14 +121,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[fetch_image_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -162,7 +158,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         client_id = get_transport_client_id(transport, client)
 
         # Kick off the conversation.
-        messages.append(
+        context.add_message(
             {
                 "role": "system",
                 "content": f"Please introduce yourself to the user. Use '{client_id}' as the user ID during function calls.",
diff --git a/examples/foundational/14d-function-calling-aws-video.py b/examples/foundational/14d-function-calling-aws-video.py
index 2b22faada..396ea0a32 100644
--- a/examples/foundational/14d-function-calling-aws-video.py
+++ b/examples/foundational/14d-function-calling-aws-video.py
@@ -101,6 +101,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         # Here we can't because AWS Bedrock doesn't support it for Claude 3.7,
         # which we need for image input.
         params=AWSBedrockLLMService.InputParams(temperature=0.8),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
     )
     llm.register_function("fetch_user_image", fetch_user_image)
 
@@ -125,14 +126,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[fetch_image_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -169,7 +163,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         client_id = get_transport_client_id(transport, client)
 
         # Kick off the conversation.
-        messages.append(
+        context.add_message(
             {
                 "role": "system",
                 "content": f"Please introduce yourself to the user. Use '{client_id}' as the user ID during function calls.",
diff --git a/examples/foundational/14d-function-calling-gemini-flash-video.py b/examples/foundational/14d-function-calling-gemini-flash-video.py
index e3d4ca8a4..c59767406 100644
--- a/examples/foundational/14d-function-calling-gemini-flash-video.py
+++ b/examples/foundational/14d-function-calling-gemini-flash-video.py
@@ -94,7 +94,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
 
     # Google Gemini model for vision analysis
-    llm = GoogleLLMService(api_key=os.getenv("GOOGLE_API_KEY"))
+    llm = GoogleLLMService(
+        api_key=os.getenv("GOOGLE_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
+    )
     llm.register_function("fetch_user_image", fetch_user_image)
 
     @llm.event_handler("on_function_calls_started")
@@ -118,14 +121,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[fetch_image_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -162,7 +158,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         client_id = get_transport_client_id(transport, client)
 
         # Kick off the conversation.
-        messages.append(
+        context.add_message(
             {
                 "role": "system",
                 "content": f"Please introduce yourself to the user. Use '{client_id}' as the user ID during function calls.",
diff --git a/examples/foundational/14d-function-calling-moondream-video.py b/examples/foundational/14d-function-calling-moondream-video.py
index 77e9282f4..d0913e3dc 100644
--- a/examples/foundational/14d-function-calling-moondream-video.py
+++ b/examples/foundational/14d-function-calling-moondream-video.py
@@ -124,7 +124,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
+    )
     llm.register_function("fetch_user_image", fetch_user_image)
 
     @llm.event_handler("on_function_calls_started")
@@ -148,14 +151,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[fetch_image_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -200,7 +196,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         client_id = get_transport_client_id(transport, client)
 
         # Kick off the conversation.
-        messages.append(
+        context.add_message(
             {
                 "role": "system",
                 "content": f"Please introduce yourself to the user. Use '{client_id}' as the user ID during function calls.",
diff --git a/examples/foundational/14d-function-calling-openai-video.py b/examples/foundational/14d-function-calling-openai-video.py
index d948e9caa..50365eb29 100644
--- a/examples/foundational/14d-function-calling-openai-video.py
+++ b/examples/foundational/14d-function-calling-openai-video.py
@@ -94,7 +94,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
+    )
     llm.register_function("fetch_user_image", fetch_user_image)
 
     @llm.event_handler("on_function_calls_started")
@@ -118,14 +121,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[fetch_image_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You are able to describe images from the user camera.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -161,7 +157,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         client_id = get_transport_client_id(transport, client)
 
         # Kick off the conversation.
-        messages.append(
+        context.add_message(
             {
                 "role": "system",
                 "content": f"Please introduce yourself to the user. Use '{client_id}' as the user ID during function calls.",
diff --git a/examples/foundational/14f-function-calling-groq.py b/examples/foundational/14f-function-calling-groq.py
index 2c922d296..e6cfb3659 100644
--- a/examples/foundational/14f-function-calling-groq.py
+++ b/examples/foundational/14f-function-calling-groq.py
@@ -67,7 +67,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = GroqLLMService(api_key=os.getenv("GROQ_API_KEY"))
+    llm = GroqLLMService(
+        api_key=os.getenv("GROQ_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
     llm.register_function("get_current_weather", fetch_weather_from_api)
@@ -93,14 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["location"],
     )
     tools = ToolsSchema(standard_tools=[weather_function])
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14g-function-calling-grok.py b/examples/foundational/14g-function-calling-grok.py
index e8529aba5..51435a2a0 100644
--- a/examples/foundational/14g-function-calling-grok.py
+++ b/examples/foundational/14g-function-calling-grok.py
@@ -67,7 +67,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = GrokLLMService(api_key=os.getenv("GROK_API_KEY"))
+    llm = GrokLLMService(
+        api_key=os.getenv("GROK_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
     llm.register_function("get_current_weather", fetch_weather_from_api)
@@ -89,14 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["location", "format"],
     )
     tools = ToolsSchema(standard_tools=[weather_function])
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14h-function-calling-azure.py b/examples/foundational/14h-function-calling-azure.py
index fb9407f1a..192cebb92 100644
--- a/examples/foundational/14h-function-calling-azure.py
+++ b/examples/foundational/14h-function-calling-azure.py
@@ -71,6 +71,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         api_key=os.getenv("AZURE_CHATGPT_API_KEY"),
         endpoint=os.getenv("AZURE_CHATGPT_ENDPOINT"),
         model=os.getenv("AZURE_CHATGPT_MODEL"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -97,14 +98,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["location", "format"],
     )
     tools = ToolsSchema(standard_tools=[weather_function])
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14i-function-calling-fireworks.py b/examples/foundational/14i-function-calling-fireworks.py
index 7a350f84f..be006e062 100644
--- a/examples/foundational/14i-function-calling-fireworks.py
+++ b/examples/foundational/14i-function-calling-fireworks.py
@@ -70,6 +70,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     llm = FireworksLLMService(
         api_key=os.getenv("FIREWORKS_API_KEY"),
         model="accounts/fireworks/models/gpt-oss-20b",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -100,14 +101,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["location", "format"],
     )
     tools = ToolsSchema(standard_tools=[weather_function])
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. Start by saying hello.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -138,6 +132,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/14k-function-calling-cerebras.py b/examples/foundational/14k-function-calling-cerebras.py
index ab340c77c..81df8723c 100644
--- a/examples/foundational/14k-function-calling-cerebras.py
+++ b/examples/foundational/14k-function-calling-cerebras.py
@@ -67,7 +67,20 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = CerebrasLLMService(api_key=os.getenv("CEREBRAS_API_KEY"))
+    llm = CerebrasLLMService(
+        api_key=os.getenv("CEREBRAS_API_KEY"),
+        system_instruction="""You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way.
+
+You have one functions available:
+
+1. get_current_weather is used to get current weather information.
+
+Infer whether to use Fahrenheit or Celsius automatically based on the location, unless the user specifies a preference.
+
+Start by asking me for my location. Then, use 'get_weather_current' to give me a forecast.
+
+    Respond to what the user said in a creative and helpful way.""",
+    )
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
     llm.register_function("get_current_weather", fetch_weather_from_api)
@@ -93,24 +106,8 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["location", "format"],
     )
     tools = ToolsSchema(standard_tools=[weather_function])
-    messages = [
-        {
-            "role": "system",
-            "content": """You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way.
 
-You have one functions available:
-
-1. get_current_weather is used to get current weather information.
-
-Infer whether to use Fahrenheit or Celsius automatically based on the location, unless the user specifies a preference.
-
-Start by asking me for my location. Then, use 'get_weather_current' to give me a forecast.
-
-    Respond to what the user said in a creative and helpful way.""",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14l-function-calling-deepseek.py b/examples/foundational/14l-function-calling-deepseek.py
index 2d54fbe07..93643146a 100644
--- a/examples/foundational/14l-function-calling-deepseek.py
+++ b/examples/foundational/14l-function-calling-deepseek.py
@@ -67,7 +67,21 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = DeepSeekLLMService(api_key=os.getenv("DEEPSEEK_API_KEY"), model="deepseek-chat")
+    llm = DeepSeekLLMService(
+        api_key=os.getenv("DEEPSEEK_API_KEY"),
+        model="deepseek-chat",
+        system_instruction="""You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way.
+
+You have one functions available:
+
+1. get_current_weather is used to get current weather information.
+
+Infer whether to use Fahrenheit or Celsius automatically based on the location, unless the user specifies a preference.
+
+Start by asking me for my location. Then, use 'get_weather_current' to give me a forecast.
+
+    Respond to what the user said in a creative and helpful way.""",
+    )
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
     llm.register_function("get_current_weather", fetch_weather_from_api)
@@ -93,24 +107,8 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["location", "format"],
     )
     tools = ToolsSchema(standard_tools=[weather_function])
-    messages = [
-        {
-            "role": "system",
-            "content": """You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way.
 
-You have one functions available:
-
-1. get_current_weather is used to get current weather information.
-
-Infer whether to use Fahrenheit or Celsius automatically based on the location, unless the user specifies a preference.
-
-Start by asking me for my location. Then, use 'get_weather_current' to give me a forecast.
-
-    Respond to what the user said in a creative and helpful way.""",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14m-function-calling-openrouter.py b/examples/foundational/14m-function-calling-openrouter.py
index 5d6589a75..9504021ff 100644
--- a/examples/foundational/14m-function-calling-openrouter.py
+++ b/examples/foundational/14m-function-calling-openrouter.py
@@ -70,7 +70,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
 
     llm = OpenRouterLLMService(
-        api_key=os.getenv("OPENROUTER_API_KEY"), model="openai/gpt-4o-2024-11-20"
+        api_key=os.getenv("OPENROUTER_API_KEY"),
+        model="openai/gpt-4o-2024-11-20",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -97,14 +99,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["location", "format"],
     )
     tools = ToolsSchema(standard_tools=[weather_function])
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14q-function-calling-qwen.py b/examples/foundational/14q-function-calling-qwen.py
index 72130b8fd..32c551b29 100644
--- a/examples/foundational/14q-function-calling-qwen.py
+++ b/examples/foundational/14q-function-calling-qwen.py
@@ -67,7 +67,11 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = QwenLLMService(api_key=os.getenv("QWEN_API_KEY"), model="qwen2.5-72b-instruct")
+    llm = QwenLLMService(
+        api_key=os.getenv("QWEN_API_KEY"),
+        model="qwen2.5-72b-instruct",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -95,14 +99,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14r-function-calling-aws.py b/examples/foundational/14r-function-calling-aws.py
index 37d29ea7f..1ce5c4220 100644
--- a/examples/foundational/14r-function-calling-aws.py
+++ b/examples/foundational/14r-function-calling-aws.py
@@ -74,6 +74,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         aws_region="us-west-2",
         model="us.anthropic.claude-haiku-4-5-20251001-v1:0",
         params=AWSBedrockLLMService.InputParams(temperature=0.8),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
     # You can also register a function_name of None to get all functions
@@ -110,14 +111,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function, restaurant_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -148,7 +142,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "user", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "user", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/14s-function-calling-sambanova.py b/examples/foundational/14s-function-calling-sambanova.py
index 76eb390c0..1548b2aa1 100644
--- a/examples/foundational/14s-function-calling-sambanova.py
+++ b/examples/foundational/14s-function-calling-sambanova.py
@@ -70,7 +70,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = SambaNovaLLMService(api_key=os.getenv("SAMBANOVA_API_KEY"))
+    llm = SambaNovaLLMService(
+        api_key=os.getenv("SAMBANOVA_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
     llm.register_function("get_current_weather", fetch_weather_from_api)
@@ -96,14 +99,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["location"],
     )
     tools = ToolsSchema(standard_tools=[weather_function])
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14t-function-calling-direct.py b/examples/foundational/14t-function-calling-direct.py
index fc12101aa..31efb7ac9 100644
--- a/examples/foundational/14t-function-calling-direct.py
+++ b/examples/foundational/14t-function-calling-direct.py
@@ -83,7 +83,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -96,14 +99,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tools = ToolsSchema(standard_tools=[get_current_weather, get_restaurant_recommendation])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14u-function-calling-ollama.py b/examples/foundational/14u-function-calling-ollama.py
index 7da07b329..b7c65dedf 100644
--- a/examples/foundational/14u-function-calling-ollama.py
+++ b/examples/foundational/14u-function-calling-ollama.py
@@ -71,7 +71,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OLLamaLLMService(model="llama3.2")  # Update to the model you're running locally
+    llm = OLLamaLLMService(
+        model="llama3.2",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )  # Update to the model you're running locally
 
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -111,14 +114,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function, restaurant_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14v-function-calling-openai.py b/examples/foundational/14v-function-calling-openai.py
index ad4aba063..920275975 100644
--- a/examples/foundational/14v-function-calling-openai.py
+++ b/examples/foundational/14v-function-calling-openai.py
@@ -78,7 +78,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
 
     # model choices: gpt-4o, gpt-4.1, etc.
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -118,14 +121,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function, restaurant_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14w-function-calling-mistral.py b/examples/foundational/14w-function-calling-mistral.py
index c62e8be05..e17d31f72 100644
--- a/examples/foundational/14w-function-calling-mistral.py
+++ b/examples/foundational/14w-function-calling-mistral.py
@@ -70,7 +70,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = MistralLLMService(api_key=os.getenv("MISTRAL_API_KEY"))
+    llm = MistralLLMService(
+        api_key=os.getenv("MISTRAL_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     # You can also register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
@@ -106,14 +109,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function, restaurant_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/14x-function-calling-openpipe.py b/examples/foundational/14x-function-calling-openpipe.py
index 576da02a4..8a71a83d7 100644
--- a/examples/foundational/14x-function-calling-openpipe.py
+++ b/examples/foundational/14x-function-calling-openpipe.py
@@ -76,6 +76,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         api_key=os.getenv("OPENAI_API_KEY"),
         openpipe_api_key=os.getenv("OPENPIPE_API_KEY"),
         tags={"conversation_id": f"pipecat-{timestamp}"},
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
     # You can also register a function_name of None to get all functions
@@ -116,14 +117,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function, restaurant_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/15-switch-voices.py b/examples/foundational/15-switch-voices.py
index 343f9faf1..4478c067b 100644
--- a/examples/foundational/15-switch-voices.py
+++ b/examples/foundational/15-switch-voices.py
@@ -112,7 +112,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = SwitchVoices()
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities. Respond to what the user said in a creative and helpful way. Your output should not include non-alphanumeric characters. You can do the following voices: 'News Lady', 'British Lady' and 'Barbershop Man'.",
+    )
     llm.register_function("switch_voice", tts.switch_voice)
 
     switch_voice_function = FunctionSchema(
@@ -128,14 +131,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[switch_voice_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities. Respond to what the user said in a creative and helpful way. Your output should not include non-alphanumeric characters. You can do the following voices: 'News Lady', 'British Lady' and 'Barbershop Man'.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -166,7 +162,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append(
+        context.add_message(
             {
                 "role": "system",
                 "content": f"Please introduce yourself to the user and let them know the voices you can do. Your initial responses should be as if you were a {tts.current_voice}.",
diff --git a/examples/foundational/15a-switch-languages.py b/examples/foundational/15a-switch-languages.py
index eb25a0511..3ad897077 100644
--- a/examples/foundational/15a-switch-languages.py
+++ b/examples/foundational/15a-switch-languages.py
@@ -103,7 +103,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = SwitchLanguage()
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities. Respond to what the user said in a creative and helpful way. Your output should not include non-alphanumeric characters. You can speak the following languages: 'English' and 'Spanish'.",
+    )
     llm.register_function("switch_language", tts.switch_language)
 
     switch_language_function = FunctionSchema(
@@ -118,14 +121,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         required=["language"],
     )
     tools = ToolsSchema(standard_tools=[switch_language_function])
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities. Respond to what the user said in a creative and helpful way. Your output should not include non-alphanumeric characters. You can speak the following languages: 'English' and 'Spanish'.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -156,7 +152,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append(
+        context.add_message(
             {
                 "role": "system",
                 "content": f"Please introduce yourself to the user and let them know the languages you speak. Your initial responses should be in {tts.current_language}.",
diff --git a/examples/foundational/16-gpu-container-local-bot.py b/examples/foundational/16-gpu-container-local-bot.py
index 93bd65200..8367e1044 100644
--- a/examples/foundational/16-gpu-container-local-bot.py
+++ b/examples/foundational/16-gpu-container-local-bot.py
@@ -70,16 +70,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         # Or, to use a local vLLM (or similar) api server
         model="meta-llama/Meta-Llama-3-8B-Instruct",
         base_url="http://0.0.0.0:8000/v1",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -111,7 +105,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     # Handle "latency-ping" messages. The client will send app messages that look like
diff --git a/examples/foundational/17-detect-user-idle.py b/examples/foundational/17-detect-user-idle.py
index e6af5a364..217162ea8 100644
--- a/examples/foundational/17-detect-user-idle.py
+++ b/examples/foundational/17-detect-user-idle.py
@@ -118,7 +118,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     llm.register_function("get_current_weather", fetch_weather_from_api)
     llm.register_function("get_restaurant_recommendation", fetch_restaurant_recommendation)
@@ -156,14 +159,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function, restaurant_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(
@@ -209,7 +205,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
         await asyncio.sleep(30)
         logger.info(f"Disabling idle detection")
diff --git a/examples/foundational/21-tavus-transport.py b/examples/foundational/21-tavus-transport.py
index 925001212..b16ebbf38 100644
--- a/examples/foundational/21-tavus-transport.py
+++ b/examples/foundational/21-tavus-transport.py
@@ -54,16 +54,12 @@ async def main():
             voice_id="a167e0f3-df7e-4d52-a9c3-f949145efdab",
         )
 
-        llm = GoogleLLMService(api_key=os.getenv("GOOGLE_API_KEY"))
+        llm = GoogleLLMService(
+            api_key=os.getenv("GOOGLE_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -95,7 +91,7 @@ async def main():
         async def on_client_connected(transport, participant):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append(
+            context.add_message(
                 {
                     "role": "system",
                     "content": "Start by greeting the user and ask how you can help.",
diff --git a/examples/foundational/21a-tavus-video-service.py b/examples/foundational/21a-tavus-video-service.py
index eff214b3a..7e32579b0 100644
--- a/examples/foundational/21a-tavus-video-service.py
+++ b/examples/foundational/21a-tavus-video-service.py
@@ -64,7 +64,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             voice_id="a167e0f3-df7e-4d52-a9c3-f949145efdab",
         )
 
-        llm = GoogleLLMService(api_key=os.getenv("GOOGLE_API_KEY"))
+        llm = GoogleLLMService(
+            api_key=os.getenv("GOOGLE_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
         tavus = TavusVideoService(
             api_key=os.getenv("TAVUS_API_KEY"),
@@ -72,14 +75,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             session=session,
         )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -113,7 +109,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append(
+            context.add_message(
                 {
                     "role": "system",
                     "content": "Start by greeting the user and ask how you can help.",
diff --git a/examples/foundational/23-bot-background-sound.py b/examples/foundational/23-bot-background-sound.py
index f2b99af19..a57e8f446 100644
--- a/examples/foundational/23-bot-background-sound.py
+++ b/examples/foundational/23-bot-background-sound.py
@@ -78,16 +78,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -128,7 +124,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         logger.info(f"Re-enabling background sound and starting bot...")
         await task.queue_frame(MixerEnableFrame(True))
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/24-user-mute-strategy.py b/examples/foundational/24-user-mute-strategy.py
index 0a1d2ffde..36b537e6a 100644
--- a/examples/foundational/24-user-mute-strategy.py
+++ b/examples/foundational/24-user-mute-strategy.py
@@ -73,7 +73,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = DeepgramTTSService(api_key=os.getenv("DEEPGRAM_API_KEY"), voice="aura-helios-en")
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful assistant who can check the weather. Always check the weather when a location is mentioned. Respond concisely and naturally. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points.",
+    )
     llm.register_function("get_current_weather", fetch_weather_from_api)
 
     weather_function = FunctionSchema(
@@ -94,14 +97,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful assistant who can check the weather. Always check the weather when a location is mentioned. Respond concisely and naturally. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(
@@ -138,7 +134,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation with a weather-related prompt
-        messages.append(
+        context.add_message(
             {
                 "role": "system",
                 "content": "Ask the user what city they'd like to know the weather for.",
diff --git a/examples/foundational/27-simli-layer.py b/examples/foundational/27-simli-layer.py
index 7c44d083a..fb469fbe1 100644
--- a/examples/foundational/27-simli-layer.py
+++ b/examples/foundational/27-simli-layer.py
@@ -68,16 +68,13 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         face_id="cace3ef7-a4c4-425d-a8cf-a5358eb0c427",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4o-mini")
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        model="gpt-4o-mini",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/28-user-assistant-turns.py b/examples/foundational/28-user-assistant-turns.py
index da38a894e..9dd1fa2a1 100644
--- a/examples/foundational/28-user-assistant-turns.py
+++ b/examples/foundational/28-user-assistant-turns.py
@@ -125,16 +125,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative, helpful, and brief way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative, helpful, and brief way. Say hello.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -169,6 +165,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Start conversation - empty prompt to let LLM follow system instructions
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/29-turn-tracking-observer.py b/examples/foundational/29-turn-tracking-observer.py
index cf85972e1..9ba22dfda 100644
--- a/examples/foundational/29-turn-tracking-observer.py
+++ b/examples/foundational/29-turn-tracking-observer.py
@@ -76,7 +76,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
     llm.register_function("get_current_weather", fetch_weather_from_api)
     llm.register_function("get_restaurant_recommendation", fetch_restaurant_recommendation)
@@ -110,14 +113,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function, restaurant_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -191,7 +187,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/30-observer.py b/examples/foundational/30-observer.py
index 02a609061..4616e0e03 100644
--- a/examples/foundational/30-observer.py
+++ b/examples/foundational/30-observer.py
@@ -107,16 +107,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -158,7 +154,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/34-audio-recording.py b/examples/foundational/34-audio-recording.py
index 46366379f..8435acbde 100644
--- a/examples/foundational/34-audio-recording.py
+++ b/examples/foundational/34-audio-recording.py
@@ -115,19 +115,16 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4")
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        model="gpt-4",
+        system_instruction="You are a helpful assistant demonstrating audio recording capabilities. Keep your responses brief and clear.",
+    )
 
     # Create audio buffer processor
     audiobuffer = AudioBufferProcessor()
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful assistant demonstrating audio recording capabilities. Keep your responses brief and clear.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/36-user-email-gathering.py b/examples/foundational/36-user-email-gathering.py
index 969a63efe..95450c695 100644
--- a/examples/foundational/36-user-email-gathering.py
+++ b/examples/foundational/36-user-email-gathering.py
@@ -79,7 +79,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     #     aiohttp_session=session,
     # )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You need to gather a valid email or emails from the user. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. If the user provides one or more email addresses confirm them with the user. Enclose all emails with <spell> tags, for example <spell>a@a.com</spell>.",
+    )
     # You can aslo register a function_name of None to get all functions
     # sent to the same callback with an additional function_name parameter.
     llm.register_function("store_user_emails", store_user_emails)
@@ -98,17 +101,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[store_emails_function])
 
-    messages = [
-        {
-            "role": "system",
-            # Cartesia <spell></spell>
-            "content": "You need to gather a valid email or emails from the user. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. If the user provides one or more email addresses confirm them with the user. Enclose all emails with <spell> tags, for example <spell>a@a.com</spell>.",
-            # Rime spell()
-            # "content": "You need to gather a valid email or emails from the user. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. If the user provides one or more email addresses confirm them with the user. Enclose all emails with spell(), for example spell(a@a.com).",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/37-mem0.py b/examples/foundational/37-mem0.py
index 423695e9c..bfc651f3a 100644
--- a/examples/foundational/37-mem0.py
+++ b/examples/foundational/37-mem0.py
@@ -221,23 +221,20 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     # )
 
     # Initialize LLM service
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4o-mini")
-
-    messages = [
-        {
-            "role": "system",
-            "content": """You are a personal assistant. You can remember things about the person you are talking to.
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        model="gpt-4o-mini",
+        system_instruction="""You are a personal assistant. You can remember things about the person you are talking to.
                         Some Guidelines:
                         - Make sure your responses are friendly yet short and concise.
                         - If the user asks you to remember something, make sure to remember it.
                         - Greet the user by their name if you know about it.
                     """,
-        },
-    ]
+    )
 
     # Set up conversation context and management
     # The context_aggregator will automatically collect conversation context
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/38-smart-turn-fal.py b/examples/foundational/38-smart-turn-fal.py
index ed4bfbae8..fe8992784 100644
--- a/examples/foundational/38-smart-turn-fal.py
+++ b/examples/foundational/38-smart-turn-fal.py
@@ -64,16 +64,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(
@@ -116,7 +112,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/38a-smart-turn-local-coreml.py b/examples/foundational/38a-smart-turn-local-coreml.py
index 4284c675c..166fbbe12 100644
--- a/examples/foundational/38a-smart-turn-local-coreml.py
+++ b/examples/foundational/38a-smart-turn-local-coreml.py
@@ -79,16 +79,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(
@@ -130,7 +126,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/38b-smart-turn-local.py b/examples/foundational/38b-smart-turn-local.py
index dc62010fb..962b608da 100644
--- a/examples/foundational/38b-smart-turn-local.py
+++ b/examples/foundational/38b-smart-turn-local.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +98,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/42-interruption-config.py b/examples/foundational/42-interruption-config.py
index 1310af908..4491f9ee0 100644
--- a/examples/foundational/42-interruption-config.py
+++ b/examples/foundational/42-interruption-config.py
@@ -62,16 +62,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(
@@ -108,7 +104,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/43-heygen-transport.py b/examples/foundational/43-heygen-transport.py
index dd31baa02..e38cb7b6e 100644
--- a/examples/foundational/43-heygen-transport.py
+++ b/examples/foundational/43-heygen-transport.py
@@ -59,16 +59,12 @@ async def main():
             voice_id="00967b2f-88a6-4a31-8153-110a92134b9f",
         )
 
-        llm = GoogleLLMService(api_key=os.getenv("GOOGLE_API_KEY"))
+        llm = GoogleLLMService(
+            api_key=os.getenv("GOOGLE_API_KEY"),
+            system_instruction="You are a helpful assistant. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Be succinct and respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful assistant. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Be succinct and respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -98,7 +94,7 @@ async def main():
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
             # Kick off the conversation.
-            messages.append(
+            context.add_message(
                 {
                     "role": "system",
                     "content": "Start by saying 'Hello' and then a short greeting.",
diff --git a/examples/foundational/43a-heygen-video-service.py b/examples/foundational/43a-heygen-video-service.py
index ba3d4417d..2b23fe1f8 100644
--- a/examples/foundational/43a-heygen-video-service.py
+++ b/examples/foundational/43a-heygen-video-service.py
@@ -66,7 +66,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             voice_id="00967b2f-88a6-4a31-8153-110a92134b9f",
         )
 
-        llm = GoogleLLMService(api_key=os.getenv("GOOGLE_API_KEY"))
+        llm = GoogleLLMService(
+            api_key=os.getenv("GOOGLE_API_KEY"),
+            system_instruction="You are a helpful assistant. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Be succinct and respond to what the user said in a creative and helpful way.",
+        )
 
         heyGen = HeyGenVideoService(
             api_key=os.getenv("HEYGEN_LIVE_AVATAR_API_KEY"),
@@ -80,14 +83,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             ),
         )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful assistant. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Be succinct and respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -131,7 +127,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
                 )
 
             # Kick off the conversation.
-            messages.append(
+            context.add_message(
                 {
                     "role": "system",
                     "content": "Start by saying 'Hello' and then a short greeting.",
diff --git a/examples/foundational/44-voicemail-detection.py b/examples/foundational/44-voicemail-detection.py
index 175cef695..f7cf14f71 100644
--- a/examples/foundational/44-voicemail-detection.py
+++ b/examples/foundational/44-voicemail-detection.py
@@ -59,19 +59,15 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
     classifier_llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
 
     voicemail = VoicemailDetector(llm=classifier_llm)
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
diff --git a/examples/foundational/45-before-and-after-events.py b/examples/foundational/45-before-and-after-events.py
index 52dfbddaf..405af49f3 100644
--- a/examples/foundational/45-before-and-after-events.py
+++ b/examples/foundational/45-before-and-after-events.py
@@ -70,16 +70,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -120,7 +116,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         # Custom frames are pushed in order so they can be used for synchronization purposes.
         await task.queue_frames([CustomBeforeProcessFrame(), LLMRunFrame(), CustomAfterPushFrame()])
 
diff --git a/examples/foundational/47-sentry-metrics.py b/examples/foundational/47-sentry-metrics.py
index e0649998d..ef20745e4 100644
--- a/examples/foundational/47-sentry-metrics.py
+++ b/examples/foundational/47-sentry-metrics.py
@@ -73,16 +73,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     llm = OpenAILLMService(
         api_key=os.getenv("OPENAI_API_KEY"),
         metrics=SentryMetrics(),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -113,7 +107,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/49a-thinking-anthropic.py b/examples/foundational/49a-thinking-anthropic.py
index 97e82666b..28dd33565 100644
--- a/examples/foundational/49a-thinking-anthropic.py
+++ b/examples/foundational/49a-thinking-anthropic.py
@@ -64,16 +64,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         params=AnthropicLLMService.InputParams(
             thinking=AnthropicLLMService.ThinkingConfig(type="enabled", budget_tokens=2048)
         ),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -104,15 +98,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append(
-            {
-                "role": "user",
-                "content": "Say hello briefly.",
-            }
-        )
+        context.add_message({"role": "user", "content": "Say hello briefly."})
         # Here are some example prompts conducive to demonstrating
         # thinking (picked from Google and Anthropic docs).
-        # messages.append(
+        # context.add_message(
         #     {
         #         "role": "user",
         #         "content": "Analogize photosynthesis and growing up. Keep your answer concise.",
diff --git a/examples/foundational/49b-thinking-google.py b/examples/foundational/49b-thinking-google.py
index 855347729..32677bcbb 100644
--- a/examples/foundational/49b-thinking-google.py
+++ b/examples/foundational/49b-thinking-google.py
@@ -69,16 +69,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
                 include_thoughts=True,
             )
         ),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -109,16 +103,11 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append(
-            {
-                "role": "user",
-                "content": "Say hello briefly.",
-            }
-        )
+        context.add_message({"role": "user", "content": "Say hello briefly."})
         # Replace the above with one of these example prompts to demonstrate
         # thinking.
         # These examples come from Gemini and Anthropic docs.
-        # messages.append(
+        # context.add_message(
         #     {
         #         "role": "user",
         #         "content": "Analogize photosynthesis and growing up. Keep your answer concise.",
diff --git a/examples/foundational/49c-thinking-functions-anthropic.py b/examples/foundational/49c-thinking-functions-anthropic.py
index 6b16a67e1..b070871b1 100644
--- a/examples/foundational/49c-thinking-functions-anthropic.py
+++ b/examples/foundational/49c-thinking-functions-anthropic.py
@@ -85,6 +85,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         params=AnthropicLLMService.InputParams(
             thinking=AnthropicLLMService.ThinkingConfig(type="enabled", budget_tokens=2048)
         ),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
     llm.register_direct_function(check_flight_status)
@@ -92,14 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tools = ToolsSchema(standard_tools=[check_flight_status, book_taxi])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -130,16 +124,11 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append(
-            {
-                "role": "user",
-                "content": "Say hello briefly.",
-            }
-        )
+        context.add_message({"role": "user", "content": "Say hello briefly."})
         # Here is an example prompt conducive to demonstrating thinking and
         # function calling.
         # This example comes from Gemini docs.
-        # messages.append(
+        # context.add_message(
         #     {
         #         "role": "user",
         #         "content": "Check the status of flight AA100 and, if it's delayed, book me a taxi 2 hours before its departure time.",
diff --git a/examples/foundational/49d-thinking-functions-google.py b/examples/foundational/49d-thinking-functions-google.py
index a377bf5ef..9b5a333de 100644
--- a/examples/foundational/49d-thinking-functions-google.py
+++ b/examples/foundational/49d-thinking-functions-google.py
@@ -90,6 +90,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
                 include_thoughts=True,
             )
         ),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
     llm.register_direct_function(check_flight_status)
@@ -97,14 +98,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tools = ToolsSchema(standard_tools=[check_flight_status, book_taxi])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages, tools)
+    context = LLMContext(tools=tools)
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -135,16 +129,11 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append(
-            {
-                "role": "user",
-                "content": "Say hello briefly.",
-            }
-        )
+        context.add_message({"role": "user", "content": "Say hello briefly."})
         # Replace the above with one of these example prompts to demonstrate
         # thinking and function calling.
         # This example comes from Gemini docs.
-        # messages.append(
+        # context.add_message(
         #     {
         #         "role": "user",
         #         "content": "Check the status of flight AA100 and, if it's delayed, book me a taxi 2 hours before its departure time.",
diff --git a/examples/foundational/52-live-translation.py b/examples/foundational/52-live-translation.py
index 861d23e37..6ab0bac18 100644
--- a/examples/foundational/52-live-translation.py
+++ b/examples/foundational/52-live-translation.py
@@ -62,16 +62,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="d4db5fb9-f44b-4bd1-85fa-192e0f0d75f9",  # Spanish-speaking Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a live translation assistant. Your sole purpose is to translate English text into Spanish. When you receive English text from the user, immediately translate it into natural, fluent Spanish. Do not add explanations, commentary, or extra information—only provide the Spanish translation of the text you receive.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a live translation assistant. Your sole purpose is to translate English text into Spanish. When you receive English text from the user, immediately translate it into natural, fluent Spanish. Do not add explanations, commentary, or extra information—only provide the Spanish translation of the text you receive.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
 
     # We use the TranscriptionUserTurnStartStrategy to start a new user turn
     # every time a transcription is received. We disable interruptions, so the
diff --git a/examples/foundational/54-context-summarization-openai.py b/examples/foundational/54-context-summarization-openai.py
index ff6701bec..25ea496c3 100644
--- a/examples/foundational/54-context-summarization-openai.py
+++ b/examples/foundational/54-context-summarization-openai.py
@@ -84,7 +84,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You have access to tools to get the current weather - use them when relevant.",
+    )
 
     # Register tool functions
     llm.register_function("get_current_weather", get_current_weather)
@@ -107,14 +110,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You have access to tools to get the current weather - use them when relevant.",
-        },
-    ]
-
-    context = LLMContext(messages, tools=tools)
+    context = LLMContext(tools=tools)
 
     # Create aggregators with summarization enabled
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
@@ -175,7 +171,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info("Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/54a-context-summarization-google.py b/examples/foundational/54a-context-summarization-google.py
index 7d2a91310..c79e96c3d 100644
--- a/examples/foundational/54a-context-summarization-google.py
+++ b/examples/foundational/54a-context-summarization-google.py
@@ -84,7 +84,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = GoogleLLMService(api_key=os.getenv("GOOGLE_API_KEY"))
+    llm = GoogleLLMService(
+        api_key=os.getenv("GOOGLE_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You have access to tools to get the current weather - use them when relevant.",
+    )
 
     # Register tool functions
     llm.register_function("get_current_weather", get_current_weather)
@@ -107,14 +110,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
     tools = ToolsSchema(standard_tools=[weather_function])
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way. You have access to tools to get the current weather - use them when relevant.",
-        },
-    ]
-
-    context = LLMContext(messages, tools=tools)
+    context = LLMContext(tools=tools)
 
     # Create aggregators with summarization enabled
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
@@ -175,7 +171,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info("Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
     @transport.event_handler("on_client_disconnected")
diff --git a/examples/foundational/55a-update-settings-deepgram-flux-stt.py b/examples/foundational/55a-update-settings-deepgram-flux-stt.py
index a482e513c..ef58fe8fb 100644
--- a/examples/foundational/55a-update-settings-deepgram-flux-stt.py
+++ b/examples/foundational/55a-update-settings-deepgram-flux-stt.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55a-update-settings-deepgram-sagemaker-stt.py b/examples/foundational/55a-update-settings-deepgram-sagemaker-stt.py
index 04451e85c..489ea1276 100644
--- a/examples/foundational/55a-update-settings-deepgram-sagemaker-stt.py
+++ b/examples/foundational/55a-update-settings-deepgram-sagemaker-stt.py
@@ -65,16 +65,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -104,7 +100,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         # NOTE: after this change, the bot will only respond if you speak Spanish
diff --git a/examples/foundational/55a-update-settings-deepgram-stt.py b/examples/foundational/55a-update-settings-deepgram-stt.py
index 8808f6f4c..b0521bb19 100644
--- a/examples/foundational/55a-update-settings-deepgram-stt.py
+++ b/examples/foundational/55a-update-settings-deepgram-stt.py
@@ -59,16 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -98,7 +94,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         # NOTE: after this change, the bot will only respond if you speak Spanish
diff --git a/examples/foundational/55b-update-settings-azure-stt.py b/examples/foundational/55b-update-settings-azure-stt.py
index 96e4041d0..1d3b3fcee 100644
--- a/examples/foundational/55b-update-settings-azure-stt.py
+++ b/examples/foundational/55b-update-settings-azure-stt.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55c-update-settings-google-stt.py b/examples/foundational/55c-update-settings-google-stt.py
index dede5b173..2956b8047 100644
--- a/examples/foundational/55c-update-settings-google-stt.py
+++ b/examples/foundational/55c-update-settings-google-stt.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55d-update-settings-assemblyai-stt.py b/examples/foundational/55d-update-settings-assemblyai-stt.py
index f57865588..f8c366a44 100644
--- a/examples/foundational/55d-update-settings-assemblyai-stt.py
+++ b/examples/foundational/55d-update-settings-assemblyai-stt.py
@@ -63,16 +63,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call demonstrating dynamic keyterms updates. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Try saying difficult names like 'Xiomara', 'Saoirse', or 'Krzystof' to test transcription accuracy.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call demonstrating dynamic keyterms updates. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Try saying difficult names like 'Xiomara', 'Saoirse', or 'Krzystof' to test transcription accuracy.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -105,7 +101,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         logger.info(
             "Phase 1: No keyterms boosting - try saying 'Xiomara', 'Saoirse', or 'Krzystof'"
         )
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(15)
diff --git a/examples/foundational/55e-update-settings-gladia-stt.py b/examples/foundational/55e-update-settings-gladia-stt.py
index e5bd5486a..5b8283f72 100644
--- a/examples/foundational/55e-update-settings-gladia-stt.py
+++ b/examples/foundational/55e-update-settings-gladia-stt.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55f-update-settings-elevenlabs-realtime-stt.py b/examples/foundational/55f-update-settings-elevenlabs-realtime-stt.py
index c3f0a6325..8d2c17778 100644
--- a/examples/foundational/55f-update-settings-elevenlabs-realtime-stt.py
+++ b/examples/foundational/55f-update-settings-elevenlabs-realtime-stt.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55g-update-settings-elevenlabs-stt.py b/examples/foundational/55g-update-settings-elevenlabs-stt.py
index 9435bc1ac..0395f273d 100644
--- a/examples/foundational/55g-update-settings-elevenlabs-stt.py
+++ b/examples/foundational/55g-update-settings-elevenlabs-stt.py
@@ -63,16 +63,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +98,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55h-update-settings-speechmatics-stt.py b/examples/foundational/55h-update-settings-speechmatics-stt.py
index c362d2f9f..ba775199a 100644
--- a/examples/foundational/55h-update-settings-speechmatics-stt.py
+++ b/examples/foundational/55h-update-settings-speechmatics-stt.py
@@ -65,16 +65,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -104,7 +100,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55i-update-settings-whisper-api-stt.py b/examples/foundational/55i-update-settings-whisper-api-stt.py
index 741601c83..962a70485 100644
--- a/examples/foundational/55i-update-settings-whisper-api-stt.py
+++ b/examples/foundational/55i-update-settings-whisper-api-stt.py
@@ -64,16 +64,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -103,7 +99,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55j-update-settings-sarvam-stt.py b/examples/foundational/55j-update-settings-sarvam-stt.py
index cab9656f8..58f0f80c7 100644
--- a/examples/foundational/55j-update-settings-sarvam-stt.py
+++ b/examples/foundational/55j-update-settings-sarvam-stt.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55k-update-settings-soniox-stt.py b/examples/foundational/55k-update-settings-soniox-stt.py
index 85b5d2ba4..b31a0e708 100644
--- a/examples/foundational/55k-update-settings-soniox-stt.py
+++ b/examples/foundational/55k-update-settings-soniox-stt.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55l-update-settings-aws-transcribe-stt.py b/examples/foundational/55l-update-settings-aws-transcribe-stt.py
index 3bfeb2faf..910bc66b4 100644
--- a/examples/foundational/55l-update-settings-aws-transcribe-stt.py
+++ b/examples/foundational/55l-update-settings-aws-transcribe-stt.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55m-update-settings-cartesia-stt.py b/examples/foundational/55m-update-settings-cartesia-stt.py
index a87847a5a..4956d8d75 100644
--- a/examples/foundational/55m-update-settings-cartesia-stt.py
+++ b/examples/foundational/55m-update-settings-cartesia-stt.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55n-update-settings-cartesia-http-tts.py b/examples/foundational/55n-update-settings-cartesia-http-tts.py
index 02d3bca2a..0423d855f 100644
--- a/examples/foundational/55n-update-settings-cartesia-http-tts.py
+++ b/examples/foundational/55n-update-settings-cartesia-http-tts.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55n-update-settings-cartesia-tts.py b/examples/foundational/55n-update-settings-cartesia-tts.py
index 04e9d8fee..2ee07b830 100644
--- a/examples/foundational/55n-update-settings-cartesia-tts.py
+++ b/examples/foundational/55n-update-settings-cartesia-tts.py
@@ -59,16 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +95,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
         # Kick off the conversation.
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55o-update-settings-elevenlabs-http-tts.py b/examples/foundational/55o-update-settings-elevenlabs-http-tts.py
index 2ca51730f..7cdf97c25 100644
--- a/examples/foundational/55o-update-settings-elevenlabs-http-tts.py
+++ b/examples/foundational/55o-update-settings-elevenlabs-http-tts.py
@@ -62,16 +62,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -101,7 +97,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55o-update-settings-elevenlabs-tts.py b/examples/foundational/55o-update-settings-elevenlabs-tts.py
index ddbfd8b8f..6d5d74a17 100644
--- a/examples/foundational/55o-update-settings-elevenlabs-tts.py
+++ b/examples/foundational/55o-update-settings-elevenlabs-tts.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id=os.getenv("ELEVENLABS_VOICE_ID"),
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55p-update-settings-openai-tts.py b/examples/foundational/55p-update-settings-openai-tts.py
index fcc24fb76..e108f4949 100644
--- a/examples/foundational/55p-update-settings-openai-tts.py
+++ b/examples/foundational/55p-update-settings-openai-tts.py
@@ -54,16 +54,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = OpenAITTSService(api_key=os.getenv("OPENAI_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -94,7 +90,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55q-update-settings-deepgram-http-tts.py b/examples/foundational/55q-update-settings-deepgram-http-tts.py
index d94bf631a..653e2bd63 100644
--- a/examples/foundational/55q-update-settings-deepgram-http-tts.py
+++ b/examples/foundational/55q-update-settings-deepgram-http-tts.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55q-update-settings-deepgram-sagemaker-tts.py b/examples/foundational/55q-update-settings-deepgram-sagemaker-tts.py
index 14958b9d2..6cd396d04 100644
--- a/examples/foundational/55q-update-settings-deepgram-sagemaker-tts.py
+++ b/examples/foundational/55q-update-settings-deepgram-sagemaker-tts.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice="aura-2-helena-en",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55q-update-settings-deepgram-tts.py b/examples/foundational/55q-update-settings-deepgram-tts.py
index e205ffa73..86d37374a 100644
--- a/examples/foundational/55q-update-settings-deepgram-tts.py
+++ b/examples/foundational/55q-update-settings-deepgram-tts.py
@@ -54,16 +54,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = DeepgramTTSService(api_key=os.getenv("DEEPGRAM_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -93,7 +89,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55r-update-settings-azure-http-tts.py b/examples/foundational/55r-update-settings-azure-http-tts.py
index 0e4df5e7c..d6b870cc7 100644
--- a/examples/foundational/55r-update-settings-azure-http-tts.py
+++ b/examples/foundational/55r-update-settings-azure-http-tts.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         region=os.getenv("AZURE_SPEECH_REGION"),
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55r-update-settings-azure-tts.py b/examples/foundational/55r-update-settings-azure-tts.py
index a32dad5ed..a564a7267 100644
--- a/examples/foundational/55r-update-settings-azure-tts.py
+++ b/examples/foundational/55r-update-settings-azure-tts.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         region=os.getenv("AZURE_SPEECH_REGION"),
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55s-update-settings-google-http-tts.py b/examples/foundational/55s-update-settings-google-http-tts.py
index ae3070124..146d5039e 100644
--- a/examples/foundational/55s-update-settings-google-http-tts.py
+++ b/examples/foundational/55s-update-settings-google-http-tts.py
@@ -54,16 +54,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = GoogleHttpTTSService(credentials=os.getenv("GOOGLE_TEST_CREDENTIALS"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -93,7 +89,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55s-update-settings-google-stream-tts.py b/examples/foundational/55s-update-settings-google-stream-tts.py
index 1aba64254..445ab537b 100644
--- a/examples/foundational/55s-update-settings-google-stream-tts.py
+++ b/examples/foundational/55s-update-settings-google-stream-tts.py
@@ -54,16 +54,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = GoogleTTSService(credentials=os.getenv("GOOGLE_TEST_CREDENTIALS"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -93,7 +89,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55u-update-settings-rime-http-tts.py b/examples/foundational/55u-update-settings-rime-http-tts.py
index 28e58ba08..f57d2b83e 100644
--- a/examples/foundational/55u-update-settings-rime-http-tts.py
+++ b/examples/foundational/55u-update-settings-rime-http-tts.py
@@ -60,16 +60,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             api_key=os.getenv("RIME_API_KEY"), voice_id="eva", aiohttp_session=session
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +95,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55u-update-settings-rime-tts.py b/examples/foundational/55u-update-settings-rime-tts.py
index 8992cb6db..5c4cdee67 100644
--- a/examples/foundational/55u-update-settings-rime-tts.py
+++ b/examples/foundational/55u-update-settings-rime-tts.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="luna",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55v-update-settings-lmnt-tts.py b/examples/foundational/55v-update-settings-lmnt-tts.py
index 01bc15ddf..0a2c27401 100644
--- a/examples/foundational/55v-update-settings-lmnt-tts.py
+++ b/examples/foundational/55v-update-settings-lmnt-tts.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="lily",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55w-update-settings-fish-tts.py b/examples/foundational/55w-update-settings-fish-tts.py
index 72a2160ba..87a4a2072 100644
--- a/examples/foundational/55w-update-settings-fish-tts.py
+++ b/examples/foundational/55w-update-settings-fish-tts.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         model="4ce7e917cedd4bc2bb2e6ff3a46acaa1",  # Barack Obama
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55x-update-settings-minimax-tts.py b/examples/foundational/55x-update-settings-minimax-tts.py
index fdb486415..66ce52071 100644
--- a/examples/foundational/55x-update-settings-minimax-tts.py
+++ b/examples/foundational/55x-update-settings-minimax-tts.py
@@ -60,16 +60,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +95,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55y-update-settings-groq-tts.py b/examples/foundational/55y-update-settings-groq-tts.py
index 3531509f2..73e4ac778 100644
--- a/examples/foundational/55y-update-settings-groq-tts.py
+++ b/examples/foundational/55y-update-settings-groq-tts.py
@@ -54,16 +54,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = GroqTTSService(api_key=os.getenv("GROQ_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -93,7 +89,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55z-update-settings-hume-tts.py b/examples/foundational/55z-update-settings-hume-tts.py
index 493550469..f16745cad 100644
--- a/examples/foundational/55z-update-settings-hume-tts.py
+++ b/examples/foundational/55z-update-settings-hume-tts.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="f898a92e-685f-43fa-985b-a46920f0650b",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55za-update-settings-neuphonic-http-tts.py b/examples/foundational/55za-update-settings-neuphonic-http-tts.py
index 6e1d18e4a..6c3e510a1 100644
--- a/examples/foundational/55za-update-settings-neuphonic-http-tts.py
+++ b/examples/foundational/55za-update-settings-neuphonic-http-tts.py
@@ -59,16 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -98,7 +94,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55za-update-settings-neuphonic-tts.py b/examples/foundational/55za-update-settings-neuphonic-tts.py
index 861167a20..e79ac666f 100644
--- a/examples/foundational/55za-update-settings-neuphonic-tts.py
+++ b/examples/foundational/55za-update-settings-neuphonic-tts.py
@@ -54,16 +54,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = NeuphonicTTSService(api_key=os.getenv("NEUPHONIC_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -93,7 +89,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zb-update-settings-inworld-http-tts.py b/examples/foundational/55zb-update-settings-inworld-http-tts.py
index 99353b87f..ccba1238c 100644
--- a/examples/foundational/55zb-update-settings-inworld-http-tts.py
+++ b/examples/foundational/55zb-update-settings-inworld-http-tts.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
         tts = InworldHttpTTSService(api_key=os.getenv("INWORLD_API_KEY"), aiohttp_session=session)
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55zb-update-settings-inworld-tts.py b/examples/foundational/55zb-update-settings-inworld-tts.py
index 104001c15..e60e65036 100644
--- a/examples/foundational/55zb-update-settings-inworld-tts.py
+++ b/examples/foundational/55zb-update-settings-inworld-tts.py
@@ -54,16 +54,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = InworldTTSService(api_key=os.getenv("INWORLD_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -93,7 +89,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zc-update-settings-gemini-tts.py b/examples/foundational/55zc-update-settings-gemini-tts.py
index 21b678047..637ee65a7 100644
--- a/examples/foundational/55zc-update-settings-gemini-tts.py
+++ b/examples/foundational/55zc-update-settings-gemini-tts.py
@@ -63,16 +63,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         ),
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +98,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zd-update-settings-aws-polly-tts.py b/examples/foundational/55zd-update-settings-aws-polly-tts.py
index 4392e7b6f..3b1c1d734 100644
--- a/examples/foundational/55zd-update-settings-aws-polly-tts.py
+++ b/examples/foundational/55zd-update-settings-aws-polly-tts.py
@@ -54,16 +54,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = AWSPollyTTSService()
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -93,7 +89,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55ze-update-settings-sarvam-http-tts.py b/examples/foundational/55ze-update-settings-sarvam-http-tts.py
index 7832a805a..ffcf086c6 100644
--- a/examples/foundational/55ze-update-settings-sarvam-http-tts.py
+++ b/examples/foundational/55ze-update-settings-sarvam-http-tts.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
         tts = SarvamHttpTTSService(api_key=os.getenv("SARVAM_API_KEY"), aiohttp_session=session)
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55ze-update-settings-sarvam-tts.py b/examples/foundational/55ze-update-settings-sarvam-tts.py
index e63c6046d..44301578f 100644
--- a/examples/foundational/55ze-update-settings-sarvam-tts.py
+++ b/examples/foundational/55ze-update-settings-sarvam-tts.py
@@ -54,16 +54,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = SarvamTTSService(api_key=os.getenv("SARVAM_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -93,7 +89,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zf-update-settings-camb-tts.py b/examples/foundational/55zf-update-settings-camb-tts.py
index 82cc4a638..f8058141d 100644
--- a/examples/foundational/55zf-update-settings-camb-tts.py
+++ b/examples/foundational/55zf-update-settings-camb-tts.py
@@ -55,16 +55,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = CambTTSService(api_key=os.getenv("CAMB_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -94,7 +90,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zg-update-settings-hathora-tts.py b/examples/foundational/55zg-update-settings-hathora-tts.py
index 80b9bfcce..2294809e2 100644
--- a/examples/foundational/55zg-update-settings-hathora-tts.py
+++ b/examples/foundational/55zg-update-settings-hathora-tts.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         model="hexgrad-kokoro-82m",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zh-update-settings-resembleai-tts.py b/examples/foundational/55zh-update-settings-resembleai-tts.py
index 44688ee25..7aba67b1f 100644
--- a/examples/foundational/55zh-update-settings-resembleai-tts.py
+++ b/examples/foundational/55zh-update-settings-resembleai-tts.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id=os.getenv("RESEMBLE_VOICE_UUID"),
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zi-update-settings-azure-llm.py b/examples/foundational/55zi-update-settings-azure-llm.py
index 43161b103..b6e8ccf71 100644
--- a/examples/foundational/55zi-update-settings-azure-llm.py
+++ b/examples/foundational/55zi-update-settings-azure-llm.py
@@ -62,16 +62,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         api_key=os.getenv("AZURE_CHATGPT_API_KEY"),
         endpoint=os.getenv("AZURE_CHATGPT_ENDPOINT"),
         model=os.getenv("AZURE_CHATGPT_MODEL"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -101,7 +95,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zi-update-settings-openai-llm.py b/examples/foundational/55zi-update-settings-openai-llm.py
index d84259cc3..c9fb026f3 100644
--- a/examples/foundational/55zi-update-settings-openai-llm.py
+++ b/examples/foundational/55zi-update-settings-openai-llm.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zj-update-settings-anthropic-llm.py b/examples/foundational/55zj-update-settings-anthropic-llm.py
index 354702880..59ac73da6 100644
--- a/examples/foundational/55zj-update-settings-anthropic-llm.py
+++ b/examples/foundational/55zj-update-settings-anthropic-llm.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = AnthropicLLMService(api_key=os.getenv("ANTHROPIC_API_KEY"))
+    llm = AnthropicLLMService(
+        api_key=os.getenv("ANTHROPIC_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zk-update-settings-google-llm.py b/examples/foundational/55zk-update-settings-google-llm.py
index cd03a34cb..4a0a3de0f 100644
--- a/examples/foundational/55zk-update-settings-google-llm.py
+++ b/examples/foundational/55zk-update-settings-google-llm.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = GoogleLLMService(api_key=os.getenv("GOOGLE_API_KEY"))
+    llm = GoogleLLMService(
+        api_key=os.getenv("GOOGLE_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zk-update-settings-google-vertex-llm.py b/examples/foundational/55zk-update-settings-google-vertex-llm.py
index 3feba582f..fe8ef808b 100644
--- a/examples/foundational/55zk-update-settings-google-vertex-llm.py
+++ b/examples/foundational/55zk-update-settings-google-vertex-llm.py
@@ -62,16 +62,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         credentials=os.getenv("GOOGLE_VERTEX_TEST_CREDENTIALS"),
         project_id=os.getenv("GOOGLE_CLOUD_PROJECT_ID"),
         location=os.getenv("GOOGLE_CLOUD_LOCATION"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -101,7 +95,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zp-update-settings-aws-bedrock-llm.py b/examples/foundational/55zp-update-settings-aws-bedrock-llm.py
index 3d3ee8fb5..72b5ae9be 100644
--- a/examples/foundational/55zp-update-settings-aws-bedrock-llm.py
+++ b/examples/foundational/55zp-update-settings-aws-bedrock-llm.py
@@ -61,16 +61,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         aws_region="us-west-2",
         model="us.anthropic.claude-haiku-4-5-20251001-v1:0",
         params=AWSBedrockLLMService.InputParams(temperature=0.8),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +94,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zq-update-settings-fal-stt.py b/examples/foundational/55zq-update-settings-fal-stt.py
index c0f0a134a..22ee38929 100644
--- a/examples/foundational/55zq-update-settings-fal-stt.py
+++ b/examples/foundational/55zq-update-settings-fal-stt.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zr-update-settings-gradium-stt.py b/examples/foundational/55zr-update-settings-gradium-stt.py
index 636d27bd8..65c70519d 100644
--- a/examples/foundational/55zr-update-settings-gradium-stt.py
+++ b/examples/foundational/55zr-update-settings-gradium-stt.py
@@ -60,16 +60,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +95,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zs-update-settings-hathora-stt.py b/examples/foundational/55zs-update-settings-hathora-stt.py
index 7a033490a..818772b75 100644
--- a/examples/foundational/55zs-update-settings-hathora-stt.py
+++ b/examples/foundational/55zs-update-settings-hathora-stt.py
@@ -61,16 +61,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +96,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zt-update-settings-nvidia-segmented-stt.py b/examples/foundational/55zt-update-settings-nvidia-segmented-stt.py
index 60a042c5f..122619aaa 100644
--- a/examples/foundational/55zt-update-settings-nvidia-segmented-stt.py
+++ b/examples/foundational/55zt-update-settings-nvidia-segmented-stt.py
@@ -57,16 +57,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -96,7 +92,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zt-update-settings-nvidia-stt.py b/examples/foundational/55zt-update-settings-nvidia-stt.py
index 415f10b12..232db9bf3 100644
--- a/examples/foundational/55zt-update-settings-nvidia-stt.py
+++ b/examples/foundational/55zt-update-settings-nvidia-stt.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zu-update-settings-openai-realtime-stt.py b/examples/foundational/55zu-update-settings-openai-realtime-stt.py
index 2bcd35f52..eb7be5c97 100644
--- a/examples/foundational/55zu-update-settings-openai-realtime-stt.py
+++ b/examples/foundational/55zu-update-settings-openai-realtime-stt.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zv-update-settings-asyncai-http-tts.py b/examples/foundational/55zv-update-settings-asyncai-http-tts.py
index 9688f1bac..79ca4382e 100644
--- a/examples/foundational/55zv-update-settings-asyncai-http-tts.py
+++ b/examples/foundational/55zv-update-settings-asyncai-http-tts.py
@@ -63,16 +63,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -102,7 +98,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55zv-update-settings-asyncai-tts.py b/examples/foundational/55zv-update-settings-asyncai-tts.py
index fe096b4be..7e5463d2b 100644
--- a/examples/foundational/55zv-update-settings-asyncai-tts.py
+++ b/examples/foundational/55zv-update-settings-asyncai-tts.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id=os.getenv("ASYNCAI_VOICE_ID", "e0f39dc4-f691-4e78-bba5-5c636692cc04"),
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zw-update-settings-gradium-tts.py b/examples/foundational/55zw-update-settings-gradium-tts.py
index d1069bfa4..e12d20dd2 100644
--- a/examples/foundational/55zw-update-settings-gradium-tts.py
+++ b/examples/foundational/55zw-update-settings-gradium-tts.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         url="wss://us.api.gradium.ai/api/speech/tts",
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zx-update-settings-cerebras-llm.py b/examples/foundational/55zx-update-settings-cerebras-llm.py
index 6123487a3..3937345e1 100644
--- a/examples/foundational/55zx-update-settings-cerebras-llm.py
+++ b/examples/foundational/55zx-update-settings-cerebras-llm.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = CerebrasLLMService(api_key=os.getenv("CEREBRAS_API_KEY"))
+    llm = CerebrasLLMService(
+        api_key=os.getenv("CEREBRAS_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zy-update-settings-deepseek-llm.py b/examples/foundational/55zy-update-settings-deepseek-llm.py
index 60cbab30b..c96107c40 100644
--- a/examples/foundational/55zy-update-settings-deepseek-llm.py
+++ b/examples/foundational/55zy-update-settings-deepseek-llm.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = DeepSeekLLMService(api_key=os.getenv("DEEPSEEK_API_KEY"))
+    llm = DeepSeekLLMService(
+        api_key=os.getenv("DEEPSEEK_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zz-update-settings-fireworks-llm.py b/examples/foundational/55zz-update-settings-fireworks-llm.py
index 97554ae19..8c4388251 100644
--- a/examples/foundational/55zz-update-settings-fireworks-llm.py
+++ b/examples/foundational/55zz-update-settings-fireworks-llm.py
@@ -61,16 +61,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     llm = FireworksLLMService(
         api_key=os.getenv("FIREWORKS_API_KEY"),
         model="accounts/fireworks/models/gpt-oss-20b",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +94,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zza-update-settings-grok-llm.py b/examples/foundational/55zza-update-settings-grok-llm.py
index 8ce081e66..e75c70b5f 100644
--- a/examples/foundational/55zza-update-settings-grok-llm.py
+++ b/examples/foundational/55zza-update-settings-grok-llm.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = GrokLLMService(api_key=os.getenv("GROK_API_KEY"))
+    llm = GrokLLMService(
+        api_key=os.getenv("GROK_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zzb-update-settings-groq-llm.py b/examples/foundational/55zzb-update-settings-groq-llm.py
index afde4499d..2f7f82bb2 100644
--- a/examples/foundational/55zzb-update-settings-groq-llm.py
+++ b/examples/foundational/55zzb-update-settings-groq-llm.py
@@ -59,17 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
 
     llm = GroqLLMService(
-        api_key=os.getenv("GROQ_API_KEY"), model="meta-llama/llama-4-maverick-17b-128e-instruct"
+        api_key=os.getenv("GROQ_API_KEY"),
+        model="meta-llama/llama-4-maverick-17b-128e-instruct",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +94,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zzc-update-settings-mistral-llm.py b/examples/foundational/55zzc-update-settings-mistral-llm.py
index 7eba98e97..ab3bd4f76 100644
--- a/examples/foundational/55zzc-update-settings-mistral-llm.py
+++ b/examples/foundational/55zzc-update-settings-mistral-llm.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = MistralLLMService(api_key=os.getenv("MISTRAL_API_KEY"))
+    llm = MistralLLMService(
+        api_key=os.getenv("MISTRAL_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zzd-update-settings-nvidia-llm.py b/examples/foundational/55zzd-update-settings-nvidia-llm.py
index ee57a3a24..a8699c71b 100644
--- a/examples/foundational/55zzd-update-settings-nvidia-llm.py
+++ b/examples/foundational/55zzd-update-settings-nvidia-llm.py
@@ -59,17 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     )
 
     llm = NvidiaLLMService(
-        api_key=os.getenv("NVIDIA_API_KEY"), model="meta/llama-3.1-405b-instruct"
+        api_key=os.getenv("NVIDIA_API_KEY"),
+        model="meta/llama-3.1-405b-instruct",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +94,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zze-update-settings-ollama-llm.py b/examples/foundational/55zze-update-settings-ollama-llm.py
index e22719ec1..4595604fd 100644
--- a/examples/foundational/55zze-update-settings-ollama-llm.py
+++ b/examples/foundational/55zze-update-settings-ollama-llm.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OLLamaLLMService(model="llama3.2")  # Update to the model you're running locally
+    llm = OLLamaLLMService(
+        model="llama3.2",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )  # Update to the model you're running locally
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zzf-update-settings-openrouter-llm.py b/examples/foundational/55zzf-update-settings-openrouter-llm.py
index fc3732192..9a381d88b 100644
--- a/examples/foundational/55zzf-update-settings-openrouter-llm.py
+++ b/examples/foundational/55zzf-update-settings-openrouter-llm.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenRouterLLMService(api_key=os.getenv("OPENROUTER_API_KEY"))
+    llm = OpenRouterLLMService(
+        api_key=os.getenv("OPENROUTER_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zzh-update-settings-qwen-llm.py b/examples/foundational/55zzh-update-settings-qwen-llm.py
index f31dc05a5..019e916cb 100644
--- a/examples/foundational/55zzh-update-settings-qwen-llm.py
+++ b/examples/foundational/55zzh-update-settings-qwen-llm.py
@@ -58,16 +58,13 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = QwenLLMService(api_key=os.getenv("QWEN_API_KEY"), model="qwen2.5-72b-instruct")
+    llm = QwenLLMService(
+        api_key=os.getenv("QWEN_API_KEY"),
+        model="qwen2.5-72b-instruct",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +94,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zzi-update-settings-sambanova-llm.py b/examples/foundational/55zzi-update-settings-sambanova-llm.py
index 96122cc03..040e95da1 100644
--- a/examples/foundational/55zzi-update-settings-sambanova-llm.py
+++ b/examples/foundational/55zzi-update-settings-sambanova-llm.py
@@ -58,16 +58,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = SambaNovaLLMService(api_key=os.getenv("SAMBANOVA_API_KEY"))
+    llm = SambaNovaLLMService(
+        api_key=os.getenv("SAMBANOVA_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -97,7 +93,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zzj-update-settings-together-llm.py b/examples/foundational/55zzj-update-settings-together-llm.py
index 710ef894a..b1c8c049d 100644
--- a/examples/foundational/55zzj-update-settings-together-llm.py
+++ b/examples/foundational/55zzj-update-settings-together-llm.py
@@ -61,16 +61,10 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     llm = TogetherLLMService(
         api_key=os.getenv("TOGETHER_API_KEY"),
         model="meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo",
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
     )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -100,7 +94,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zzl-update-settings-nvidia-tts.py b/examples/foundational/55zzl-update-settings-nvidia-tts.py
index a8bd50dcd..5bc89bd58 100644
--- a/examples/foundational/55zzl-update-settings-nvidia-tts.py
+++ b/examples/foundational/55zzl-update-settings-nvidia-tts.py
@@ -55,16 +55,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
 
     tts = NvidiaTTSService(api_key=os.getenv("NVIDIA_API_KEY"))
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -94,7 +90,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/55zzm-update-settings-speechmatics-tts.py b/examples/foundational/55zzm-update-settings-speechmatics-tts.py
index 39ed792dd..d9eaaec86 100644
--- a/examples/foundational/55zzm-update-settings-speechmatics-tts.py
+++ b/examples/foundational/55zzm-update-settings-speechmatics-tts.py
@@ -59,16 +59,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
             aiohttp_session=session,
         )
 
-        llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+        llm = OpenAILLMService(
+            api_key=os.getenv("OPENAI_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -98,7 +94,9 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         @transport.event_handler("on_client_connected")
         async def on_client_connected(transport, client):
             logger.info(f"Client connected")
-            messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+            context.add_message(
+                {"role": "system", "content": "Please introduce yourself to the user."}
+            )
             await task.queue_frames([LLMRunFrame()])
 
             await asyncio.sleep(10)
diff --git a/examples/foundational/55zzn-update-settings-groq-stt.py b/examples/foundational/55zzn-update-settings-groq-stt.py
index b00ecda81..1d488119c 100644
--- a/examples/foundational/55zzn-update-settings-groq-stt.py
+++ b/examples/foundational/55zzn-update-settings-groq-stt.py
@@ -60,16 +60,12 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
         voice_id="71a7ad14-091c-4e8e-a314-022ece01c121",  # British Reading Lady
     )
 
-    llm = OpenAILLMService(api_key=os.getenv("OPENAI_API_KEY"))
+    llm = OpenAILLMService(
+        api_key=os.getenv("OPENAI_API_KEY"),
+        system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+    )
 
-    messages = [
-        {
-            "role": "system",
-            "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-        },
-    ]
-
-    context = LLMContext(messages)
+    context = LLMContext()
     user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
         context,
         user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -99,7 +95,7 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
     @transport.event_handler("on_client_connected")
     async def on_client_connected(transport, client):
         logger.info(f"Client connected")
-        messages.append({"role": "system", "content": "Please introduce yourself to the user."})
+        context.add_message({"role": "system", "content": "Please introduce yourself to the user."})
         await task.queue_frames([LLMRunFrame()])
 
         await asyncio.sleep(10)
diff --git a/examples/foundational/56-lemonslice-transport.py b/examples/foundational/56-lemonslice-transport.py
index 8b2e19a6e..a3b6b60d3 100644
--- a/examples/foundational/56-lemonslice-transport.py
+++ b/examples/foundational/56-lemonslice-transport.py
@@ -55,21 +55,17 @@ async def main():
 
         stt = DeepgramSTTService(api_key=os.getenv("DEEPGRAM_API_KEY"))
 
-        llm = GroqLLMService(api_key=os.getenv("GROQ_API_KEY"))
+        llm = GroqLLMService(
+            api_key=os.getenv("GROQ_API_KEY"),
+            system_instruction="You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
+        )
 
         tts = ElevenLabsTTSService(
             api_key=os.getenv("ELEVENLABS_API_KEY", ""),
             voice_id=os.getenv("ELEVENLABS_VOICE_ID", ""),
         )
 
-        messages = [
-            {
-                "role": "system",
-                "content": "You are a helpful LLM in a WebRTC call. Your goal is to demonstrate your capabilities in a succinct way. Your output will be spoken aloud, so avoid special characters that can't easily be spoken, such as emojis or bullet points. Respond to what the user said in a creative and helpful way.",
-            },
-        ]
-
-        context = LLMContext(messages)
+        context = LLMContext()
         user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
             context,
             user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
@@ -101,7 +97,7 @@ async def main():
         async def on_client_connected(transport, participant):
             logger.info("Client connected")
             # Kick off the conversation.
-            messages.append(
+            context.add_message(
                 {
                     "role": "system",
                     "content": "Start by greeting the user and ask how you can help.",
diff --git a/src/pipecat/services/anthropic/llm.py b/src/pipecat/services/anthropic/llm.py
index 03190ef99..755ca6400 100644
--- a/src/pipecat/services/anthropic/llm.py
+++ b/src/pipecat/services/anthropic/llm.py
@@ -17,7 +17,7 @@ import io
 import json
 import re
 from dataclasses import dataclass, field
-from typing import Any, ClassVar, Dict, List, Literal, Optional, Union
+from typing import Any, Dict, List, Literal, Optional, Union
 
 import httpx
 from loguru import logger
@@ -38,7 +38,6 @@ from pipecat.frames.frames import (
     LLMFullResponseEndFrame,
     LLMFullResponseStartFrame,
     LLMMessagesFrame,
-    LLMTextFrame,
     LLMThoughtEndFrame,
     LLMThoughtStartFrame,
     LLMThoughtTextFrame,
@@ -219,6 +218,7 @@ class AnthropicLLMService(LLMService):
         client=None,
         retry_timeout_secs: Optional[float] = 5.0,
         retry_on_timeout: Optional[bool] = False,
+        system_instruction: Optional[str] = None,
         **kwargs,
     ):
         """Initialize the Anthropic LLM service.
@@ -230,6 +230,7 @@ class AnthropicLLMService(LLMService):
             client: Optional custom Anthropic client instance.
             retry_timeout_secs: Request timeout in seconds for retry logic.
             retry_on_timeout: Whether to retry the request once if it times out.
+            system_instruction: Optional system instruction to use as the system prompt.
             **kwargs: Additional arguments passed to parent LLMService.
         """
         params = params or AnthropicLLMService.InputParams()
@@ -265,6 +266,9 @@ class AnthropicLLMService(LLMService):
         )  # if the client is provided, use it and remove it, otherwise create a new one
         self._retry_timeout_secs = retry_timeout_secs
         self._retry_on_timeout = retry_on_timeout
+        self._system_instruction = system_instruction
+        if self._system_instruction:
+            logger.debug(f"{self}: Using system instruction: {self._system_instruction}")
 
     def can_generate_metrics(self) -> bool:
         """Check if this service can generate usage metrics.
@@ -395,9 +399,11 @@ class AnthropicLLMService(LLMService):
         # Universal LLMContext
         if isinstance(context, LLMContext):
             adapter: AnthropicLLMAdapter = self.get_llm_adapter()
-            params = adapter.get_llm_invocation_params(
+            params: AnthropicLLMInvocationParams = adapter.get_llm_invocation_params(
                 context, enable_prompt_caching=self._settings.enable_prompt_caching
             )
+            if self._system_instruction:
+                params["system"] = self._system_instruction
             return params
 
         # Anthropic-specific context
diff --git a/src/pipecat/services/aws/llm.py b/src/pipecat/services/aws/llm.py
index 540ac4a8e..31f8085bc 100644
--- a/src/pipecat/services/aws/llm.py
+++ b/src/pipecat/services/aws/llm.py
@@ -19,7 +19,7 @@ import json
 import os
 import re
 from dataclasses import dataclass, field
-from typing import Any, ClassVar, Dict, List, Optional
+from typing import Any, Dict, List, Optional
 
 from loguru import logger
 from PIL import Image
@@ -39,7 +39,6 @@ from pipecat.frames.frames import (
     LLMFullResponseEndFrame,
     LLMFullResponseStartFrame,
     LLMMessagesFrame,
-    LLMTextFrame,
     UserImageRawFrame,
 )
 from pipecat.metrics.metrics import LLMTokenUsage
@@ -781,6 +780,7 @@ class AWSBedrockLLMService(LLMService):
         client_config: Optional[Config] = None,
         retry_timeout_secs: Optional[float] = 5.0,
         retry_on_timeout: Optional[bool] = False,
+        system_instruction: Optional[str] = None,
         **kwargs,
     ):
         """Initialize the AWS Bedrock LLM service.
@@ -795,6 +795,7 @@ class AWSBedrockLLMService(LLMService):
             client_config: Custom boto3 client configuration.
             retry_timeout_secs: Request timeout in seconds for retry logic.
             retry_on_timeout: Whether to retry the request once if it times out.
+            system_instruction: Optional system instruction to use as the system prompt.
             **kwargs: Additional arguments passed to parent LLMService.
         """
         params = params or AWSBedrockLLMService.InputParams()
@@ -840,8 +841,11 @@ class AWSBedrockLLMService(LLMService):
 
         self._retry_timeout_secs = retry_timeout_secs
         self._retry_on_timeout = retry_on_timeout
+        self._system_instruction = system_instruction
 
         logger.info(f"Using AWS Bedrock model: {model}")
+        if self._system_instruction:
+            logger.debug(f"{self}: Using system instruction: {self._system_instruction}")
 
     def can_generate_metrics(self) -> bool:
         """Check if the service can generate usage metrics.
@@ -1019,7 +1023,9 @@ class AWSBedrockLLMService(LLMService):
         # Universal LLMContext
         if isinstance(context, LLMContext):
             adapter: AWSBedrockLLMAdapter = self.get_llm_adapter()
-            params = adapter.get_llm_invocation_params(context)
+            params: AWSBedrockLLMInvocationParams = adapter.get_llm_invocation_params(context)
+            if self._system_instruction:
+                params["system"] = [{"text": self._system_instruction}]
             return params
 
         # AWS Bedrock-specific context
diff --git a/src/pipecat/services/openai/base_llm.py b/src/pipecat/services/openai/base_llm.py
index 40a2672f8..ed012d0df 100644
--- a/src/pipecat/services/openai/base_llm.py
+++ b/src/pipecat/services/openai/base_llm.py
@@ -11,7 +11,7 @@ import base64
 import json
 from contextlib import asynccontextmanager
 from dataclasses import dataclass, field
-from typing import Any, ClassVar, Dict, List, Mapping, Optional
+from typing import Any, Dict, List, Mapping, Optional
 
 import httpx
 from loguru import logger
@@ -117,6 +117,7 @@ class BaseOpenAILLMService(LLMService):
         params: Optional[InputParams] = None,
         retry_timeout_secs: Optional[float] = 5.0,
         retry_on_timeout: Optional[bool] = False,
+        system_instruction: Optional[str] = None,
         **kwargs,
     ):
         """Initialize the BaseOpenAILLMService.
@@ -131,6 +132,7 @@ class BaseOpenAILLMService(LLMService):
             params: Input parameters for model configuration and behavior.
             retry_timeout_secs: Request timeout in seconds. Defaults to 5.0 seconds.
             retry_on_timeout: Whether to retry the request once if it times out.
+            system_instruction: Optional system instruction to prepend to messages.
             **kwargs: Additional arguments passed to the parent LLMService.
         """
         params = params or BaseOpenAILLMService.InputParams()
@@ -155,6 +157,7 @@ class BaseOpenAILLMService(LLMService):
         )
         self._retry_timeout_secs = retry_timeout_secs
         self._retry_on_timeout = retry_on_timeout
+        self._system_instruction = system_instruction
         self._full_model_name: str = ""
         self._client = self.create_client(
             api_key=api_key,
@@ -165,6 +168,9 @@ class BaseOpenAILLMService(LLMService):
             **kwargs,
         )
 
+        if self._system_instruction:
+            logger.debug(f"{self}: Using system instruction: {self._system_instruction}")
+
     def create_client(
         self,
         api_key=None,
@@ -285,6 +291,14 @@ class BaseOpenAILLMService(LLMService):
         params.update(params_from_context)
 
         params.update(self._settings.extra)
+
+        # Prepend system instruction if set
+        if self._system_instruction:
+            messages = params.get("messages", [])
+            params["messages"] = [
+                {"role": "system", "content": self._system_instruction}
+            ] + messages
+
         return params
 
     async def run_inference(