Swap to gpt oss (#19)

Switched to gpt model family and added eval metric (#18)

Updated evaluation report

Increase threshold, fix eval prompt

Improved eval set

Deploy Fixes and clean up (#14)

Reduce top k pick (#10)

Added observability via LangFuse (#9)

Use versioned prompt

Fixed dynamic model load (#8)

Add rate-limit handling, checkpointing, and saving deepeval report (#5)

Improved ingestion, retrieve and generate (#4)

RAG evaluation and prompts