feat: update codebase to use httpx2 and enhance docstrings for clarity

app/services/llm.py

@@ -1,25 +1,28 @@
-from collections.abc import AsyncGenerator
+"""Streaming client for OpenAI-compatible chat completion endpoints."""
 
-import httpx
+from collections.abc import AsyncIterator
+
+import httpx2
 import orjson
 
 
 class StreamLLMService:
+    """Send prompts to an LLM endpoint and yield newline-delimited JSON events."""
+
     def __init__(self, base_url: str = "http://localhost:11434/v1") -> None:
+        """Configure the endpoint and default model."""
         self.base_url = base_url
         self.model = "llama3.2"
 
-    async def stream_chat(self, prompt: str) -> AsyncGenerator[bytes]:
-        """Stream chat completion responses from LLM."""
-        # Send the user a message first
+    async def stream_chat(self, prompt: str) -> AsyncIterator[bytes]:
+        """Yield the prompt followed by streamed model response chunks."""
         user_msg = {
             "role": "user",
             "content": prompt,
         }
         yield orjson.dumps(user_msg) + b"\n"
 
-        # Open client as context manager and stream responses
-        async with httpx.AsyncClient(base_url=self.base_url) as client:
+        async with httpx2.AsyncClient(base_url=self.base_url) as client:
             async with client.stream(
                 "POST",
                 "/chat/completions",
@@ -32,21 +35,16 @@ class StreamLLMService:
             ) as response:
                 async for line in response.aiter_lines():
                     if line.startswith("data: ") and line != "data: [DONE]":
-                        try:
-                            json_line = line[6:]  # Remove "data: " prefix
-                            data = orjson.loads(json_line)
-                            content = (
-                                data.get("choices", [{}])[0]
-                                .get("delta", {})
-                                .get("content", "")
-                            )
-                            if content:
-                                model_msg = {"role": "model", "content": content}
-                                yield orjson.dumps(model_msg) + b"\n"
-                        except Exception:
-                            pass
+                        data = orjson.loads(line.removeprefix("data: "))
+                        choices = data.get("choices", [])
+                        if not choices:
+                            continue
+                        content = choices[0].get("delta", {}).get("content", "")
+                        if content:
+                            model_msg = {"role": "model", "content": content}
+                            yield orjson.dumps(model_msg) + b"\n"
 
 
-# FastAPI dependency
 def get_llm_service(base_url: str | None = None) -> StreamLLMService:
+    """Build the service using the supplied URL or its local default."""
     return StreamLLMService(base_url=base_url or "http://localhost:11434/v1")