itzcrazykns/vane

Vane is a privacy-focused, AI-powered search and answering engine you can run entirely on your own computer or server. Think of it as a self-hosted alternative to services like Perplexity, it takes your questions, searches the web, and returns a summarized answer with cited sources, all without sending your queries to any external company. The core idea is simple: when you type a question, Vane runs a web search in the background using SearxNG (a privacy-respecting meta-search engine that queries multiple search engines simultaneously without tracking you), collects the relevant results, and feeds them to an AI language model that synthesizes a coherent answer. Because the whole pipeline runs locally, your search history stays on your machine. Every search is also saved locally so you can revisit past research anytime. You can plug in the AI model of your choice. If you prefer to keep everything offline, it supports Ollama to run local models on your own hardware. If you are comfortable using cloud-based AI, it also connects to OpenAI, Anthropic Claude, Google Gemini, and Groq. You can even switch between a "Speed Mode" for quick answers and a "Quality Mode" for deeper research that takes a bit longer. Beyond plain text search, Vane supports image and video search, file uploads (so you can ask questions about a PDF or image you provide), domain-specific search (restricting results to certain websites), contextual widgets for things like weather or stock prices, and a browsable "Discover" section for trending content. You would reach for Vane if you want the convenience of an AI-powered search assistant but are uncomfortable with commercial services logging your queries. It is particularly useful for researchers, developers doing technical lookups, or anyone privacy-conscious who wants full control over their AI stack. The project is built with TypeScript and runs as a web app, deployable in minutes via Docker.