Bug: Unhandled LLM RuntimeError crashes /api/chat with raw 500 if Ollama/Groq goes down

[anushagarg6]

Description

If Ollama crashes or Groq rate-limits mid-request, ask_ollama() in backend/llm/client.py raises a RuntimeError. This exception is not caught in the /api/chat route in backend/app.py, so it propagates as an unhandled 500 Internal Server Error with a raw Python traceback.

Reproduction

1. Start the backend
2. Stop Ollama (pkill ollama) or exhaust Groq rate limits
3. Send any chat message
4. Observe: raw 500 error, no user-friendly message

Impact

Users see a confusing error with no indication of what went wrong or how to fix it. The frontend displays a generic failure with no actionable guidance.

Fix Needed

Wrap the answer_question() call in /api/chat with a try/except RuntimeError block:

try:
    result = answer_question(question, session_id, language)
except RuntimeError as e:
    raise HTTPException(status_code=503, detail=str(e))

This returns a clean 503 Service Unavailable with the specific LLM error message (e.g. "Cannot connect to Ollama. Is 'ollama serve' running?").

Files Affected

• backend/app.py — /api/chat route