Vane

Vane is an open-source, self-hosted AI answering engine that combines web search with large language models. It functions as a privacy-focused interface that delivers answers with cited sources, allowing users to maintain control over their search data by running the software on their own hardware.

The application is deployed as a web server. While Docker is the recommended installation method for simplified setup, it can also be installed manually using npm. It can be run as a bundled container including its own search engine or connected to an existing external instance. Users interact with the software through a web browser to perform research, analyze documents, and browse trending content.

Key features

• Support for local LLMs via Ollama and cloud providers including OpenAI, Claude, and Gemini
• Search modes including Speed, Balanced, and Quality for different research depths
• Web search integration powered by SearxNG
• Ability to upload PDFs, text files, and images for document-based questioning
• Domain-specific searching to limit results to particular websites
• UI widgets for weather, calculations, and stock prices
• Integrated image and video search capabilities
• Local storage of search history for revisiting previous discoveries

Architecturally, Vane relies on SearxNG to aggregate search results while remaining agnostic regarding the AI backend. This allows users to integrate local LLMs through Ollama or connect to cloud providers such as OpenAI, Anthropic Claude, Google Gemini, and Groq. The software supports a slim installation for those who already manage a SearxNG instance, provided that the instance has JSON format and the Wolfram Alpha search engine enabled. It is designed for researchers and privacy-conscious users who require the capabilities of a generative AI search engine without relying on centralized platforms.

Vane is positioned as a private, self-hosted tool for AI-assisted web exploration and document analysis.