Steadlith
Stable Incremental Indexing for Changing RAG Corpora
A Python toolkit for stable, incremental RAG indexing that combines content-defined chunking, content-addressed identities, cache-aware planning, and transactional indexing so small edits produce small index updates.
Built using Python, RAG, SQLite, Content-Defined Chunking
- Plans chunk additions, moves, keeps, and deletions before applying changes
- Reuses embeddings through stable content identities and a transactional cache
- Includes deterministic benchmarks for chunk churn and retrieval quality
If you're curious, feel free to explore: