Vane
Vane, formerly Perplexica, is an MIT-licensed AI answering engine that runs on your own hardware. It searches the web through SearXNG, which its Docker image includes, and answers with cited sources, using local models through Ollama or hosted models from OpenAI, Anthropic, Google and Groq. It has speed, balanced and quality modes, searches discussions and academic papers, reads uploaded files, and needs no paid search API.
What does Vane do?
Vane is "a privacy-focused AI answering engine that runs entirely on your own hardware." The repository was renamed from Perplexica to Vane, and GitHub redirects the old address. Web search is "powered by SearxNG", a self-hosted metasearch engine, and answers come "with cited sources." Models run locally through Ollama or come from OpenAI, Anthropic Claude, Google Gemini, Groq and others, including any local OpenAI-compatible server.
There are three modes, speed, balanced, and quality for deep research. You can search the web, discussions or academic papers, limit a search to named domains, and upload PDFs, text files and images. Image and video search, widgets for weather, calculations and stock prices, search suggestions and a discover feed are included, and history is saved locally. It has an API for developers and can be set as the browser's search engine. Authentication and custom agents are listed as upcoming, and Tavily and Exa search as coming soon.
Key facts
- Licence: MIT, copyright ItzCrazyKns, with no added conditions.
- Renamed from Perplexica to Vane; GitHub redirects ItzCrazyKns/Perplexica to ItzCrazyKns/Vane.
- Web search through SearXNG, bundled in the Docker image; no paid search API is required.
- Models: local through Ollama, Lemonade or any OpenAI-compatible server; hosted from OpenAI, Anthropic, Google Gemini and Groq.
- Speed, balanced and quality modes; sources from the web, discussions or academic papers; domain-limited search.
- File uploads (PDF, text, images); image and video search; widgets; a discover feed; local search history; a developer API.
- Install: docker run of itzcrazykns1337/vane on port 3000, a slim image for your own SearXNG, or npm with a separate SearXNG.
- Latest tagged release when read: v1.12.2 on 2026-04-10; last push 2026-09-01; 36,866 stars on 2026-09-17.
What does it replace, and where does it fall short?
Vane replaces Perplexity with the same shape, a question in and a cited answer out, running on your own machine with a model you choose and a search engine you host, and with no search API bill. It is the closest open-source match on the open-source alternatives to Perplexity page.
Where it falls short: there is no sign-in or multi-user support yet, which the README lists as upcoming, and no mobile or desktop app. Search quality depends on SearXNG and the model you run. The last tagged release was April 2026, though the repository had pushes in September. For sign-in and a choice of search providers, Morphic.
How does Vane run?
Run the Docker image on port 3000; it includes SearXNG, so the README says no additional setup is required. Enter model keys or a local Ollama address on the setup screen. A slim image works with your own SearXNG, which must have JSON output and the Wolfram Alpha engine enabled; without Docker, npm install, build and start alongside a separate SearXNG. The maintainer sells no hosted version; third-party one-click deploys are listed.
Who is Vane for?
Someone who wants a private Perplexity on their own server or laptop with local models and no search bill. A team that needs user accounts should look at Morphic, which has authentication.
What limits does the README state?
From the README: Tavily and Exa support is "coming soon"; more sources and integrations are "in progress"; authentication is an unchecked upcoming feature; on Linux, Ollama needs OLLAMA_HOST set to 0.0.0.0:11434 to be reached from Docker.
Questions people ask
Is Vane the same as Perplexica?
Yes. The repository was renamed from Perplexica to Vane, and GitHub redirects ItzCrazyKns/Perplexica to ItzCrazyKns/Vane. The README now uses the name Vane throughout, and only two one-click deploy links still mention Perplexica. It is the same MIT project.
Does Vane need a paid search API?
No. Web search runs through SearXNG, a self-hosted metasearch engine that the Docker image includes. Tavily and Exa are listed as coming soon as optional providers. You still need a model: a hosted API key or a local model through Ollama.
Can Vane run with local models only?
Yes. It supports Ollama, Lemonade and any local OpenAI-compatible server, and with SearXNG bundled in the same container the search and the model both stay on your hardware. Answers still come from live web results fetched by SearXNG.
How does Vane compare with Perplexity?
Vane gives the same cited-answer shape on your own hardware, with the model and search engine under your control and no subscription. Perplexity is a hosted service with its own index and apps. The open-source alternatives to Perplexity page compares Vane with Morphic and Khoj.
Does Vane support multiple users?
Not yet. Authentication is listed in the README under upcoming features, unchecked. For now it is suited to one person or a trusted network. Morphic supports user sign-in through Supabase.
Sources
- Vane README and MIT LICENSE: github.com/ItzCrazyKns/Vane, read 2026-09-17.
Compared with the others
On the open-source alternatives to Perplexity page, Vane is the pick for the closest perplexity shape, self-hosted. MIT, cited answers from web search through a bundled SearXNG, local models through Ollama or hosted ones, speed, balanced and quality modes, and file uploads.
Also on that page: Morphic for a richer answer page with accounts, Khoj for answers from your documents as well as the web.
More answer engines
An AGPL-3.0 personal AI that answers from the web and your own documents with any local or hosted model, builds custom agents, sends scheduled research by email, and runs self-hosted or in its maintainers' cloud.
An Apache-2.0 AI search engine with a generative interface that answers with citations, using OpenAI, Anthropic, Google, Ollama or any OpenAI-compatible model and Tavily, SearXNG, Brave or Exa for search, with sign-in and shareable results.
Added September 17, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 17, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/ItzCrazyKns/Vane.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.