chunktuner¶

Give it your documents. It tries multiple chunking strategies, measures which setup supports retrieval best, and recommends a configuration for your corpus and use case.

What it does¶

Chunking choices directly affect RAG quality. chunktuner benchmarks strategies (fixed windows, recursive splits, semantic splits, PDF structure, code AST, and more), scores them with retrieval metrics (token recall, MRR, NDCG) and optional generation metrics (RAGAS), then surfaces a winner.

your docs → try multiple strategies → measure each → recommend the best config

Three ways to use it¶

Install¶

uv (tool)¶

uv tool install chunktuner

pip¶

pip install chunktuner

Library only¶

uv add chunktuner

Where to go next¶

• Quickstart — install, first commands, minimal Python example
• Strategy guide — choosing a strategy
• CLI reference — every chunk-tune command
• Python API — library patterns
• API reference — auto-generated module docs
• MCP setup — Claude Desktop and CHUNK_TUNER_BASE_DIR
• Contributing — how to contribute (links to repo checklist)