aiplanethub/beyondllm

Provides modular RAG pipeline components (data ingestion from YouTube/PDFs, retrieval, LLM generation) with pluggable embeddings and LLM models from OpenAI, Google, and open-source providers. Includes built-in evaluation metrics for retrieval quality (hit rate, MRR) and generation quality (context relevancy, answer relevancy, groundedness), plus cost and latency monitoring for closed-source models via the Observer component.

292 stars. No commits in the last 6 months.