Local Deep Research

Local Deep Research is an open-source, self-hosted web application that functions as an AI research assistant. It performs agentic searches across the web and private documents to synthesize information into reports with citations. The software is deployed as a server, typically via Docker or pip, and integrates with LLM providers and search engines to automate the collection of data.

Users can conduct research using different strategies, ranging from quick pipelines for fast facts to autonomous agents for complex academic analysis. The application allows users to build a personal knowledge base by downloading sources from research sessions, such as academic papers and web pages, which are then indexed and made searchable for future queries. This allows knowledge to compound over time as the user adds more documents to their local library.

Key features

• Autonomous agentic research mode using LangGraph
• Integration with academic engines including arXiv, PubMed, and Semantic Scholar
• Encrypted personal libraries for storing and indexing downloaded sources
• Per-user AES-256 encrypted databases via SQLCipher
• Support for multiple LLM endpoints including Ollama and OpenAI
• Multiple research modes including Quick Summary and Report Generation
• Authenticated REST API for external integration
• PDF export for generated research reports

The system architecture emphasizes data sovereignty and privacy. It contains no telemetry, no analytics, and no tracking, ensuring that all usage metrics stay in the local encrypted database. Each user is provided with an isolated SQLCipher database, and the application does not store passwords, instead using them to decrypt the database at login. It is designed for researchers, academics, and individuals who require a private environment for knowledge management and retrieval-augmented generation. The software is compatible with x86-64 CPUs with AVX support and ARM64 architectures.

Local Deep Research is positioned as a privacy-focused tool for deep analysis and personal knowledge compounding.