feat(chat): add optional response format parameter to request_chat_completion

python/ebook_search/llm_interface.py

@@ -174,6 +174,8 @@ async def request_chat_completion(
     client: httpx.AsyncClient,
     config: EbookSearchConfig,
     messages: Sequence[dict[str, str]],
+    *,
+    response_format: dict[str, object] | None = None,
 ) -> str:
     """Request a chat completion over a shared async client.
 
@@ -181,6 +183,7 @@ async def request_chat_completion(
         client (httpx.AsyncClient): Shared async client whose connection pool bounds concurrency.
         config (EbookSearchConfig): Runtime settings supplying the endpoint, model, and auth.
         messages (Sequence[dict[str, str]]): OpenAI-style chat messages.
+        response_format (dict[str, object] | None): Optional OpenAI-compatible structured output constraint.
 
     Returns:
         str: The assistant message text.
@@ -192,11 +195,8 @@ async def request_chat_completion(
         response = await client.post(
             f"{config.vllm_base_url.rstrip('/')}/chat/completions",
             headers=auth_headers(config.vllm_api_key),
-            json={
-                "model": config.chat_model,
-                "messages": list(messages),
-                "temperature": 0,
-            },
+            json={"model": config.chat_model, "messages": list(messages), "temperature": 0}
+            | ({"response_format": response_format} if response_format is not None else {}),
             timeout=config.chat_timeout_seconds,
         )
         response.raise_for_status()