khoj-ai/khoj

Khoj is an open-source, self-hostable AI second brain that turns your personal documents into a searchable, conversational knowledge base. It uses retrieval-augmented generation (RAG) to answer questions by combining your local files — PDFs, Markdown, Org mode, and more — with real-time web search results. You can run it entirely on your own hardware for full data privacy, or connect it to cloud LLM providers like OpenAI, Anthropic, or Google. What makes Khoj worth the 33,000+ stars on GitHub is its flexibility. It supports over 20 LLM backends including Llama 3, GPT, Claude, Gemini, DeepSeek, Qwen, and Mistral. You access it through whatever interface fits your workflow — browser, Obsidian plugin, Emacs client, desktop app, mobile, or WhatsApp. Custom agents let you build specialized assistants with their own knowledge bases and tool access. Scheduled automations handle recurring research tasks and push notifications when results are ready. Recent additions include Pipali, an open-source desktop AI co-worker, and a v2.0 beta with significant architecture improvements. The project ships 170+ releases under AGPL-3.0 with an active community on Discord.