Vane

What is Vane?

Vane is an open source AI answering engine that searches the web and uses a large language model to write answers with cited sources. It combines a SearXNG metasearch backend with your choice of LLM, and runs entirely on your own hardware so queries never leave your network. Vane is the renamed successor to the Perplexica project by the same author.

What is Vane best for?

Vane suits anyone who wants Perplexity-style cited AI search without sending every question to a hosted cloud — privacy-conscious users, homelabbers, and teams that want to point the tool at local models through Ollama. It fits best when you already run (or don’t mind running) SearXNG and a model provider.

What can Vane do?

• Answer questions from live web search with inline source citations (RAG architecture)
• Use local models through Ollama or connect to OpenAI, Anthropic Claude, Google Gemini, Groq, and other OpenAI-compatible APIs
• Switch between Speed, Balanced, and Quality search modes to trade latency for depth
• Pull from multiple source types — web, discussions, academic papers, images, and videos
• Upload and query documents, including PDFs, text files, and images
• Show quick-answer widgets for weather, calculations, and stock prices
• Limit a search to specific domains and keep a local search history

Where does Vane fall short?

Vane is still under active development, and features the README lists as upcoming — such as built-in authentication and custom agents — aren’t there yet, so it’s not ready as a shared multi-user service out of the box. It also isn’t a standalone product: answer quality depends entirely on the LLM and the SearXNG instance you wire up, so a weak local model or a rate-limited metasearch backend produces weak answers. And unlike hosted Perplexity, there’s no polished mobile app or account sync — you get a web UI you host yourself.

Is Vane free?

Yes — Vane is free and open source under the MIT license, with no paid tier or hosted plan. Self-hosting the software costs nothing. Your only costs are optional: if you connect a commercial LLM API instead of a local model, you pay that provider per token. Running a local model through Ollama keeps it fully free.

What does Vane replace?

Vane is a self-hosted stand-in for hosted AI answer engines — a Perplexity alternative first and foremost, and it covers the web-search-and-answer use case people reach ChatGPT and Google Gemini for. The trade is control and privacy for the convenience of a managed cloud app.

FAQ

Is Vane open source? Yes. Vane is released under the MIT license, one of the most permissive open source licenses, and the full source is on GitHub. It’s the rebranded continuation of the earlier Perplexica project.

Can I self-host Vane for free? Yes. Vane is free to self-host, with a one-line Docker image (itzcrazykns1337/vane) as the recommended path. Pair it with a local model through Ollama and there are no per-query costs at all.

What do I need to run Vane? A machine that runs Docker, a SearXNG instance (Vane can bundle one or use your own), and an LLM provider — either a local model via Ollama or an API key for OpenAI, Claude, Gemini, or another OpenAI-compatible service.

Is Vane a good Perplexity alternative? For privacy and control, yes — it delivers cited web-search answers on your own infrastructure with your own models. It trades Perplexity’s polish, mobile apps, and zero-setup convenience for a self-hosted stack you assemble and maintain.