Add Request Timeout + Proper Error Handling to Ollama Calls

[Mario5T]

Description

Currently, the application makes requests to the Ollama server without any timeout, retry logic, or structured error handling. This can cause the process to hang indefinitely if the server is unresponsive and lacks resilience against transient network anomalies.
This issue implements production-critical reliability improvements to the textToJSON generation loop by introducing:

• A configurable request timeout (default 30s)
• A retry mechanism with a backoff/limit (default 3 retries)
• Proper exception handling using requests.exceptions.Timeout and RequestException
• Graceful error messaging to debug networking and resilience issues
  Making this change ensures defensive programming patterns are established without altering the overall architecture.

[lonewolve24]

Hello @marcvergees has this issue been solved would like to work on it

[Mahendrareddy2006]

Hi! @Mario5T @marcvergees I was exploring the repository while preparing for GSoC and noticed this issue about improving reliability of the Ollama requests.

In my recent PR #271 , I added a configurable request timeout to the requests.post() call in src/llm.py to prevent the process from hanging indefinitely if the Ollama server becomes unresponsive.

This partially addresses the concern raised here regarding missing timeout handling. Additional improvements like retry logic and backoff could still be added on top of that if needed.

Happy to help extend the implementation with retries or additional error handling if that would be useful.

[PriyadharshiniVenkatraj]

Hi! I was going through this issue and noticed that a timeout was recently introduced in another PR, but retry logic and more robust error handling are still not fully implemented.

I’d like to work on extending this by:

• adding retry logic with exponential backoff
• handling different request exceptions (timeouts, connection errors, etc.)
• ensuring clearer and structured error messages for debugging

If someone is already actively working on the retry/backoff part, I’m happy to collaborate or take on any remaining pieces. Otherwise, I’d love to proceed with a PR.

Please let me know your thoughts!