An AI-powered research assistant that performs deep, agentic research using multiple LLMs and search engines with full local deployment and encryption.
Local Deep Research is an AI-powered research assistant that automates deep, multi-source investigation and builds a personal, encrypted knowledge base. It solves the problem of time-consuming manual research by automatically searching academic papers, web sources, and private documents, then synthesizing findings into coherent reports with citations. The project emphasizes privacy, allowing everything to run locally with encrypted data storage.
Researchers, academics, journalists, and developers who need to perform thorough, cited research on complex topics while maintaining data privacy and control. It's also suitable for enterprises wanting to integrate AI research capabilities with their existing knowledge bases.
Developers choose Local Deep Research for its strong privacy guarantees, local deployment options, and flexibility in LLM and search engine selection. Its unique selling point is the combination of agentic research automation with a fully encrypted, user-owned knowledge base that compounds research over time.
Local Deep Research achieves ~95% on SimpleQA benchmark (tested with GPT-4.1-mini). Supports local and cloud LLMs (Ollama, Google, Anthropic, ...). Searches 10+ sources - arXiv, PubMed, web, and your private documents. Everything Local & Encrypted.
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Automatically searches across 10+ sources including arXiv, PubMed, and the web, synthesizing findings into reports with citations, as highlighted in the README's key features.
Uses per-user SQLCipher databases with AES-256 encryption and has no telemetry, ensuring complete data control and privacy, as stated in the security section.
Supports both local models via Ollama and cloud providers like OpenAI, allowing users to choose based on cost and privacy needs, evidenced in the LLM support section.
Connects to existing vector stores via LangChain retrievers, enabling seamless integration with FAISS, Chroma, and other databases, as described in the enterprise features.
Requires configuring multiple components like Docker, Ollama, and SearXNG, which can be time-consuming and error-prone for non-technical users, as shown in the quick start instructions.
Benchmark accuracy is ~95% but preliminary and depends heavily on model choice and configuration, meaning results may not be consistently reliable for critical research.
Admits that credentials are held in plain text in process memory during sessions, a vulnerability noted in the security transparency section, despite industry-wide acceptance.
Running local LLMs requires significant GPU or CPU resources, and cloud model usage can incur costs, making it less suitable for budget-constrained projects.