Anthropic LLM service

src/dailyai/services/anthropic_llm_service.py

@@ -0,0 +1,51 @@
+import asyncio
+import os
+from typing import AsyncGenerator
+from anthropic import AsyncAnthropic
+from dailyai.pipeline.frames import Frame, LLMMessagesQueueFrame, TextFrame
+
+from dailyai.services.ai_services import LLMService
+
+
+class AnthropicLLMService(LLMService):
+
+    def __init__(self, api_key, model="claude-3-opus-20240229", max_tokens=1024):
+        super().__init__()
+        self.client = AsyncAnthropic(api_key=api_key)
+        self.model = model
+        self.max_tokens = max_tokens
+
+    async def process_frame(self, frame: Frame) -> AsyncGenerator[Frame, None]:
+        if not isinstance(frame, LLMMessagesQueueFrame):
+            yield frame
+
+        stream = await self.client.messages.create(
+            max_tokens=self.max_tokens,
+            messages=[
+                {
+                    "role": "user",
+                    "content": "Hello, Claude",
+                }
+            ],
+            model=self.model,
+            stream=True,
+        )
+        async for event in stream:
+            if event.type == "content_block_delta":
+                yield TextFrame(event.delta.text)
+
+
+async def test():
+    service = AnthropicLLMService(api_key=os.getenv("ANTHROPIC_API_KEY"))
+    messages = [
+        {
+            "role": "user",
+            "content": "Hello, Claude",
+        }
+    ]
+    async for frame in service.process_frame(LLMMessagesQueueFrame(messages)):
+        print(frame)
+
+
+if __name__=="__main__":
+    asyncio.run(test())