Storia-AI/sage

Supports pluggable retrieval strategies ranging from lightweight LLM-only approaches to full RAG pipelines with custom embeddings and vector stores (Marqo, OpenAI, Anthropic). Indexes codebases for semantic search and context-aware Q&A, with documented benchmarks comparing different retrieval and embedding strategies. Modular architecture allows swapping LLM providers, embeddings, and vector store backends through abstract class implementations.

1,265 stars. No commits in the last 6 months.