feat: set store=False and add run_inference tests

Set store=False in Responses API calls since we send full conversation
history as input items and don't use previous_response_id.

Add 5 run_inference tests for OpenAIResponsesLLMService using real
LLMContext and adapter (only HTTP client mocked).

src/pipecat/adapters/services/open_ai_responses_adapter.py

@@ -79,9 +79,19 @@ class OpenAIResponsesLLMAdapter(BaseLLMAdapter[OpenAIResponsesLLMInvocationParam
             # message when instructions are provided. Contexts that worked with
             # OpenAILLMService (system_instruction + empty messages) need the
             # instructions converted to an initial developer message.
-            # NOTE: once we support `previous_response_id`, we need to revisit
+            #
+            # NOTE: if/when we support `previous_response_id`, we'll need to revisit
             # this logic, as it'll be legit to provide instructions without input
-            # items if `previous_response_id` is provided.
+            # items if `previous_response_id` is provided. Though...OpenAI's docs +
+            # ChatGPT suggests that `previous_response_id` is primarily for
+            # development convenience, not performance (other than minor bandwidth
+            # savings from not transferring the full context), as the model still
+            # processes the full context from the previous response. The tradeoff
+            # of using `previous_response_id` is that it requires enabling OpenAI-side
+            # 30-day conversation storage (meaning we couldn't do `store=False`
+            # in the API call), which may not be desirable for all users. So,
+            # my guess is we won't need to support `previous_response_id` in the
+            # immediate future.
             if not input_items:
                 params["input"] = [{"role": "developer", "content": system_instruction}]
             else:

src/pipecat/services/openai/responses/llm.py

@@ -324,6 +324,7 @@ class OpenAIResponsesLLMService(LLMService):
         params: Dict[str, Any] = {
             "model": self._settings.model,
             "stream": True,
+            "store": False,
             "input": invocation_params["input"],
         }
 

tests/test_run_inference.py

@@ -801,6 +801,7 @@ async def test_openai_responses_run_inference_with_llm_context():
         call_kwargs = service._client.responses.create.call_args.kwargs
         assert call_kwargs["model"] == "gpt-4.1"
         assert call_kwargs["stream"] is False
+        assert call_kwargs["store"] is False
         assert call_kwargs["input"] == [{"role": "user", "content": "Hello, world!"}]
         assert call_kwargs["instructions"] == "You are a helpful assistant"
         assert call_kwargs["temperature"] == 0.7