modelscope/sirchmunk

Health v2 · maintenance only

Adoption not part of the score

1333 stars · 147 forks observed · 2026-09-22

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Sirchmunk is an agentic, embedding-free search engine that turns raw files into a self-evolving knowledge base in real time, without vector databases or pre-indexing. It combines Monte Carlo evidence sampling, knowledge clustering, and a ReAct agent fallback, exposed via MCP, REST, WebSocket, CLI, and a Web UI.

Use cases

• search raw files instantly without building a vector index
• build a self-evolving knowledge base from documents
• replace brittle embedding-based RAG pipelines
• query large document collections token-efficiently with an LLM
• add agentic search to AI assistants via MCP
• chat with my files in real time

When to choose

• your data changes frequently and static embeddings go stale
• you want to skip ETL and vector database setup entirely
• you need precise evidence extraction from raw files rather than chunk similarity
• you want an MCP-integrated search backend for agents

When to avoid

• you need sub-millisecond retrieval over massive fixed corpora where a tuned vector index excels
• you want fully offline search with no LLM API dependency
• you need a mature, battle-tested production search engine
• your workload requires heavy multi-modal media processing beyond text extraction

Facets

application · maturity active

search-engine rag agent-framework mcp chat-interface llm-inference nlp artificial-intelligence large-language-models developer-tools python self-hosted cli cross-platform agentic-search embedding-free indexless-retrieval self-evolving-knowledge monte-carlo-sampling react-agent fastapi nextjs duckdb knowledge-base retrieval-augmented-generation search ai-agents docker web-server

Member repositories

For agents

markdown · JSON · MCP: product_card(name="modelscope/sirchmunk")