Vane (formerly Perplexica)
Self-run AI answering engine that searches via SearXNG and answers with cited sources, using local or cloud LLMs.
Vane, formerly Perplexica, answers questions with cited sources using SearXNG and your choice of LLM.
Vane, the project formerly called Perplexica, is an AI answering engine that runs on your own hardware. It searches through SearXNG, then has a local or cloud model write an answer with cited sources. It suits researchers and privacy-minded users who want a Perplexity-style tool they control.
What it is
You run one Docker image that bundles Vane with SearXNG, open it in the browser and configure your AI provider. It offers Speed, Balanced and Quality search modes, source choices for web, discussions and academic papers, file uploads for questions about documents, and image and video search.
Who it's for
It suits individual researchers, students and technical users who are happy to run a container and supply their own model or API key. It does not suit you if you want a managed service with guaranteed answer quality, since results depend on the SearXNG engines and model you connect.
Key features
Cited answers
Answers come with sources so you can check where claims came from.
Three search modes
Speed, Balanced and Quality modes trade response time against depth of research.
Choose your sources
Search the web, discussions or academic papers, or limit a search to specific domains.
Bundled SearXNG
The Docker image includes SearXNG, so a separate search backend is not required to start.
Document questions
Upload PDFs, text files and images and ask questions about them.
Local or cloud models
Use Ollama locally or connect OpenAI, Claude, Gemini, Groq and others, mixing models as you like.
Up and running in 3 steps
Our honest verdict
What's great
- One docker run command starts the app with its search engine included, then you finish setup in a browser screen.
- A slim image lets you point Vane at a SearXNG instance you already operate.
- Widgets show weather, calculations and stock prices as cards when relevant.
- Search history is stored locally so you can revisit earlier research.
Where it falls short
- You must run Docker and supply your own LLM, either a local model through Ollama or a paid API key, so quality and cost depend on that choice.
- Without Docker you must install SearXNG yourself, enable JSON format and Wolfram Alpha, then build with npm.
- The README says Tavily and Exa support is still coming, so search quality rests on SearXNG results for now.
- The project was renamed from Perplexica to Vane, so older tutorials, image names and links may not match.
Pricing & license
Vane is free under the MIT licence and the project shows no paid plan. The only cost is running it yourself, plus any LLM API charges if you use a cloud model.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Is Perplexica the same as Vane?
Yes. The GitHub repository now uses the name Vane and the old Perplexica address redirects to it; the README title and Docker image use Vane.
Does it work fully offline?
Searching the web needs internet, but you can pair it with local models through Ollama so your prompts are not sent to a cloud LLM provider.
What does it cost?
The software is free under MIT. You pay for any cloud model API you connect and for the machine that runs the container.
