Raw Data to Self-Evolving Intelligence
Sirchmunk is an open-source, embedding-free, agentic search engine that transforms raw data into self-evolving intelligence in real time.
- Embedding-Free Retrieval — No vector database, no pre-indexing, no ETL pipeline. Drop your files and search immediately.
- Self-Evolving Knowledge — Knowledge clusters compound with every search, delivering faster and richer results over time.
- Monte Carlo Evidence Sampling — Extract precise evidence from large documents without reading them in full.
- ReAct Agent Fallback — An autonomous agent iteratively explores alternative strategies when standard retrieval falls short.
- Multi-Surface Integration — MCP, REST API, WebSocket, CLI, and Web UI — all built in.
- Token-Efficient Design — LLM inference only when necessary. Monte Carlo sampling and knowledge reuse minimize costs.
- Source Code: github.com/modelscope/sirchmunk
- Documentation: modelscope.github.io/sirchmunk-web/docs/
- Blog & Technical Report: modelscope.github.io/sirchmunk-web/blog/
- Releases: github.com/modelscope/sirchmunk/releases
Apache License 2.0