Update llm_handler.py

llm_handler.py

@@ -1,6 +1,7 @@
 # rentbot/llm_handler.py
 import os
 from openai import AsyncOpenAI
+import json
 
 client = AsyncOpenAI(api_key=os.getenv("OPENAI_API_KEY"))
 
@@ -34,10 +35,11 @@ tools = [
     }
 ]
 
-async def get_llm_response(messages: list):
+async def get_llm_response(messages: list, async_chunk_handler):
     """
-    Yields response chunks from OpenAI's chat completion streaming API.
-    Returns the full assistant message and any tool calls.
+    Calls the OpenAI API and streams text chunks to a handler.
+    This is now a regular async function, NOT a generator.
+    It returns the final assistant message and any tool calls.
     """
     try:
         stream = await client.chat.completions.create(
@@ -53,33 +55,33 @@ async def get_llm_response(messages: list):
         
         async for chunk in stream:
             delta = chunk.choices[0].delta
-            if delta.content:
+            if delta and delta.content:
                 text_chunk = delta.content
                 full_response += text_chunk
-                yield text_chunk # Yield text chunks for real-time TTS
+                # Call the provided handler with the new chunk
+                await async_chunk_handler(text_chunk)
             
-            if delta.tool_calls:
-                # Accumulate tool call chunks
+            if delta and delta.tool_calls:
+                # This part handles accumulating tool call data from multiple chunks
                 if not tool_calls:
-                    tool_calls.extend(delta.tool_calls)
-                else:
-                    for i, tool_call_chunk in enumerate(delta.tool_calls):
-                        if tool_call_chunk.function.arguments:
-                            tool_calls[i].function.arguments += tool_call_chunk.function.arguments
+                    tool_calls = [{"id": tc.id, "type": "function", "function": {"name": None, "arguments": ""}} for tc in delta.tool_calls]
+                
+                for i, tool_call_chunk in enumerate(delta.tool_calls):
+                    if tool_call_chunk.function.name:
+                        tool_calls[i]["function"]["name"] = tool_call_chunk.function.name
+                    if tool_call_chunk.function.arguments:
+                        tool_calls[i]["function"]["arguments"] += tool_call_chunk.function.arguments
 
         # Construct the final assistant message object
         assistant_message = {"role": "assistant", "content": full_response}
         if tool_calls:
-            assistant_message["tool_calls"] = [
-                {
-                    "id": tc.id,
-                    "type": "function",
-                    "function": {"name": tc.function.name, "arguments": tc.function.arguments}
-                } for tc in tool_calls
-            ]
+            assistant_message["tool_calls"] = tool_calls
             
+        # This return is now VALID because there is no 'yield' in this function
         return assistant_message, tool_calls
 
     except Exception as e:
         print(f"Error in get_llm_response: {e}")
-        return {"role": "assistant", "content": "I'm having a little trouble right now. Please try again in a moment."}, []
+        error_message = "I'm having a little trouble right now. Please try again in a moment."
+        await async_chunk_handler(error_message)
+        return {"role": "assistant", "content": error_message}, []