fix: use max_completion_tokens and remove temperature for Azure OpenAI compat

- Replace max_tokens with max_completion_tokens (required by newer Azure models)
- Remove hardcoded temperature (not supported by all model types)
- Add response body logging on LLM API errors for easier debugging

backend/routes/ask.py

@@ -204,18 +204,18 @@ async def _call_llm(question: str, events: list[dict]) -> str:
     else:
         headers["Authorization"] = f"Bearer {LLM_API_KEY}"
 
+    payload = {
+        "model": LLM_MODEL,
+        "messages": messages,
+        "max_completion_tokens": 800,
+    }
+
     async with httpx.AsyncClient(timeout=LLM_TIMEOUT_SECONDS) as client:
-        resp = await client.post(
+        resp = await client.post(url, headers=headers, json=payload)
-            url,
+        if resp.status_code >= 400:
-            headers=headers,
+            body = resp.text
-            json={
+            logger.error("LLM API error", status_code=resp.status_code, url=url, response_body=body)
-                "model": LLM_MODEL,
+            raise RuntimeError(f"LLM API error {resp.status_code}: {body[:500]}")
-                "messages": messages,
-                "temperature": 0.3,
-                "max_tokens": 800,
-            },
-        )
-        resp.raise_for_status()
         data = resp.json()
         return data["choices"][0]["message"]["content"].strip()