backend/main.py relies on the LLM returning valid JSON every time. A small retry loop (1–2 retries) with a stricter prompt on failure, or switching to a JSON schema–enforced generation call, would eliminate a potential failure point. (This seems like a lower priority item given LLMs use JSON as their native language).
backend/main.pyrelies on the LLM returning valid JSON every time. A small retry loop (1–2 retries) with a stricter prompt on failure, or switching to a JSON schema–enforced generation call, would eliminate a potential failure point. (This seems like a lower priority item given LLMs useJSONas their native language).