Skip to content
Get packs

Guide · Open source

How to use Vane: install with Docker, connect models and search

Updated · written with AI from the project's own documentation and checked against our catalogue data

Vane is an AI answering engine you run on your own machine. You ask a question, it searches the web through SearxNG, and a language model you choose writes an answer with cited sources. This guide shows you how to start Vane with Docker, connect a model, and avoid the most common setup problems.

VaneItzCrazyKns/Vane · ★ 38k · MIT

Quick start

  1. 1Install Docker. Make sure Docker is installed and running on your computer.
  2. 2Start Vane. Run the docker run command from this guide to pull and start the Vane container with bundled SearxNG.
  3. 3Open the setup screen. Go to http://localhost:3000 in your browser.
  4. 4Add a model provider. Enter your API keys or local model settings, such as an Ollama URL, in the setup screen.
  5. 5Ask a question. Type a question and review the answer and its cited sources.

What you need before you install Vane

The main prerequisite for the recommended setup is Docker. Vane also needs a language model to produce answers, so you need either a local model served through Ollama or an account with a cloud provider such as OpenAI, Claude, or Groq.

Vane is a self-hosted tool. The README centers on running it yourself, so it is not a zero-setup hosted service.

  • Docker, installed and running on your system
  • A model provider: a local model through Ollama, or an API key from OpenAI, Claude, or Groq

Install and start Vane with Docker

Run this single command to pull and start Vane. The image includes the bundled SearxNG search engine, so no separate search setup is required.

Once the container is running, open http://localhost:3000 in your browser. The README says you can then configure your settings, such as API keys and models, directly in the setup screen. The -v flag creates a persistent volume for your data and uploaded files.

docker run -d -p 3000:3000 -v vane-data:/home/vane/data --name vane itzcrazykns1337/vane:latest

Configure a model and run your first search

In the setup screen at http://localhost:3000, add the AI provider you want to use. Vane supports local LLMs through Ollama and cloud providers including OpenAI, Anthropic Claude, Google Gemini, and Groq. You can mix and match models depending on the task.

After setup, type a question into the search box. Vane searches the web and writes an answer with cited sources, which you can check by following the links.

  • Pick a search mode: Speed Mode for quick answers, Balanced Mode for everyday searches, or Quality Mode for deeper research.
  • Choose a source type: web, discussions, or academic papers.
  • Use the search history feature to revisit earlier searches, which are saved locally.

Connect Vane to Ollama for local models

If you want to use local models through Ollama, set the Ollama API URL in Vane's settings menu. The URL depends on your operating system. On Windows and Mac, the README uses http://host.docker.internal:11434. On Linux, use http://<private_ip_of_host>:11434, replacing the placeholder with your host's private IP address.

On Linux, the README says Ollama must be reachable from the network. Add Environment="OLLAMA_HOST=0.0.0.0:11434" inside /etc/systemd/system/ollama.service, reload with systemctl daemon-reload, and restart with systemctl restart ollama. Also make sure your firewall does not block port 11434.

systemctl daemon-reload systemctl restart ollama

Use your own SearxNG instance or build from source

If you already run SearxNG, the README describes a slim image that connects to your instance through the SEARXNG_API_URL variable. Your SearxNG instance must have JSON format enabled and the Wolfram Alpha search engine enabled.

To build from source, clone the repository, build the image with Docker, and run it. The README notes that Docker is the recommended route, and that a non-Docker install is also described, which requires installing SearXNG first and running npm commands.

  • Slim image: set SEARXNG_API_URL to your SearxNG URL and use the slim-latest tag.
  • Build from source: clone the repository, then build and run the image as shown in the README.
  • Non-Docker install: install SearXNG, clone the repository, run npm i, npm run build, and npm run start.
git clone https://github.com/ItzCrazyKns/Vane.git

What people use it for

  • Asking a research question and checking the cited web sources behind the answer
  • Searching academic papers or discussions instead of only general web pages
  • Limiting a search to specific websites when you know where to look, such as technical documentation
  • Uploading a document such as a PDF and asking questions about it
  • Running local models through Ollama so questions are answered on your own hardware

Frequently asked questions

Is Vane free to use?

Vane is released under the MIT licence, and the source code is on GitHub. The README does not list prices, but you may still pay for cloud model providers such as OpenAI, Claude, or Groq if you use them.

What do I need to run Vane?

The README lists Docker as the prerequisite for the recommended install. You also need a language model, either a local one through Ollama or a cloud provider account, because Vane needs a model to write answers.

Does Vane work with Ollama?

Yes. The README says Vane can use local LLMs through Ollama. Set the Ollama API URL in Vane's settings menu, using the address that matches your operating system.

Why does Vane say Ollama connection error?

The README says this usually means the backend cannot reach Ollama's API. Check the API URL in settings, and on Linux make sure Ollama listens on 0.0.0.0:11434 and that your firewall allows port 11434.

How is Vane different from Perplexica or Open WebUI?

The README does not compare Vane with these projects, so this guide cannot make a claim about them. Vane is described as an AI answering engine that searches through SearxNG and cites sources. Compare the projects' own documentation for details.

Does Vane support Tavily or Exa search?

Not yet. The README says support for Tavily and Exa search is coming soon, so do not rely on them for now.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Vane suits people who want AI answers with cited sources, who are comfortable running Docker, and who want to use local models through Ollama or their own SearxNG setup. If you want a zero-setup hosted service, or you do not have a model provider or local LLM, pick a different tool.

Install options & downloads