monkeSearch

monkeSearch is an open-source, offline-first desktop search tool that allows you to find local files using natural language. Instead of relying on simple keyword matching, it uses embedding models to understand the semantic meaning of your query, allowing you to search for concepts, not just words. We made it Temporal Aware, such that it can parse time-related queries like "documents from last week" and apply them as filters. The entire process, from indexing to searching, runs locally on your machine. Our project utilizes different backends optimized for each OS to deliver the best performance and user experience across platforms.

Implementation Versions: There are multiple implementations in different branches written to achieve the same task for testing purposes. Rigorous evaluations and testing will be done before finalizing a single one for the main release.

For Agentic Use: The legacy LLM-based implementations (branches below) are particularly suitable for integration into larger AI pipelines and agentic systems. These versions allow direct filesystem access through natural language without modifying any files, leveraging OS-level scoped safety through Spotlight. If you're building autonomous agents or LLM orchestration systems that need file discovery capabilities, these branches provide a direct LLM-to-filesystem bridge without the overhead of maintaining a separate index.

• Initial implementation using LangExtract (Legacy – LLM based, ideal for agentic pipelines)
• Current main branch: LEANN-based (except Windows: Windows uses ChromaDB) semantic search with temporal awareness via regex parsing
• llama.cpp rewrite (legacy-main-llm-implementation) – deprecated but useful for direct LLM integration
• llama.cpp feature branch – another variation of the llama.cpp rewrite, with a detailed response model