monkeSearch

Spec sheet

Purpose

monkeSearch is an offline-first desktop tool for searching local files with natural language.Read 11 Oct 2026 · monkesearch.github.io

Product

monkeSearch is an open-source, offline-first desktop tool for searching local files with natural-language queries.Read 5 Oct 2026 · monkesearch.github.io

Integrations

The main/dev approach uses macOS Spotlight and is described as a natural-language filesystem bridge for AI pipelines and agents.Read 11 Oct 2026 · monkesearch.github.io

Security

The maker describes the Spotlight-based approach as read-only and scoped through Spotlight’s access controls.Read 5 Oct 2026 · monkesearch.github.io

Support

The project page asks users who find the open-source project useful to star its GitHub repository.Read 11 Oct 2026 · monkesearch.github.io

License

The repository lists the project under the Apache-2.0 license.Read 11 Oct 2026 · github.com

Agent use

The maker says the LLM-based implementation can serve as a natural-language file-discovery bridge for AI agents and orchestration systems.Read 5 Oct 2026 · monkesearch.github.io

Benchmark

The maker's benchmark page says its query-parsing evaluation covers 80 queries across file type, temporal awareness, combined queries, and noise resistance.Read 11 Oct 2026 · monkesearch.github.io

Current platform status

The repository describes the main branch as currently macOS-only and says cross-platform support is in the vectordb branch.Read 11 Oct 2026 · github.com

Data handling

The maker says nothing leaves the machine; the Spotlight approach is read-only and uses Spotlight's access controls.Read 11 Oct 2026 · monkesearch.github.io

License and support

The project is open source under the Apache-2.0 license, and its site asks users to support it by starring the repository.Read 5 Oct 2026 · github.com

Limitations

The repository says the main implementation searches Spotlight-indexed files, searches metadata rather than file contents, and supports only simple time expressions.Read 11 Oct 2026 · github.com

Local processing

The project says both its LLM and vector database approaches run fully offline.Read 11 Oct 2026 · monkesearch.github.io

macOS approach

The main/dev branch uses a local LLM to turn queries into macOS Spotlight search arguments.Read 11 Oct 2026 · monkesearch.github.io

macOS integration

The main/dev branch uses Spotlight’s existing index and the NSMetadataQuery API for file discovery.Read 5 Oct 2026 · monkesearch.github.io

Performance tradeoff

On macOS, the maker says disabling LEANN recompute makes searches faster at the cost of disk space; recompute is enabled by default.Read 5 Oct 2026 · monkesearch.github.io

Platform architecture

The maker describes vectordb support for macOS, Windows, and Linux; the repository says the main branch is currently macOS-only.Read 5 Oct 2026 · github.com

Requirements

The repository lists macOS, Python 3.8+, llama-server from llama.cpp, and a GGUF model as requirements.Read 11 Oct 2026 · github.com

Search inputs

The maker describes queries as combining file type, temporal context, and other keywords.Read 5 Oct 2026 · monkesearch.github.io

Search queries

Queries can combine file type, temporal context, and other keywords.Read 11 Oct 2026 · monkesearch.github.io

Vector database approach

The vectordb branch uses vector databases instead of a live LLM at query time and is described as cross-platform.Read 11 Oct 2026 · monkesearch.github.io

Windows integration

The vectordb Windows architecture uses an os.walk filesystem crawl and ChromaDB for vector storage and retrieval.Read 5 Oct 2026 · monkesearch.github.io

Where Specifiction places it

• Desktop Search Software#33 of 577.2