local-deep-research

Local Deep Research is a self-hosted, privacy-focused AI research assistant that conducts multi-step agentic research by combining any supported LLM with more than ten search backends, including arXiv, PubMed, SearXNG, Brave, and local private document collections. Users interact through a web UI served on localhost, deployable via Docker, Docker Compose, or pip install. The system connects to Claude through Anthropic's API alongside other cloud providers and local inference runtimes such as Ollama and llama.cpp, treating all as interchangeable LLM backends. Research history and knowledge bases are stored in a SQLCipher-encrypted SQLite database, meaning all data stays on the user's hardware. A concrete benchmark result distinguishes the project: running Qwen3.6-27B on a single RTX 3090, it achieved roughly 95% accuracy on SimpleQA (n=500), claimed as the first fully local open-source system to reach that mark. Researchers, academics, and privacy-conscious developers who need cited, multi-source reports without sending queries to third-party services are the primary audience.

Lo que la gente pregunta sobre local-deep-research