Open-Source AI Search
Read the code. Understand the system. Adapt an MIT-licensed search engine to your own workflow.
Built in the open Open Source Self-Hosted AI Search
Vane: The Open-Source Perplexity Alternative
Meet Vane, formerly Perplexica. Explore an open-source AI search engine with cited answers, flexible model support, and self-hosting options.
An independent community guide. Not the official Vane website.
THE SEARCH WORKFLOW
A question worth exploring.↵
Search the webSearXNG · multiple search engines
01Connect your modelLocal with Ollama · or a cloud provider
02An answer. With sources.
[1] Original source[2] Read furtherDiscover. Compare. Self-host.Open-source AI search, on your terms.
Vane is an open-source AI search engine for people who want more control over how they find information. Instead of stopping at a list of links, it combines web results with a language model to produce a readable answer with citations. The project was previously called Perplexica and continues under the Vane name.
Its search layer uses SearXNG, a metasearch engine that gathers results from multiple search engines. A model then turns retrieved context into an answer. You can connect a local model through Ollama or choose a supported cloud provider, depending on your hardware, budget, and preferences.
It is a useful starting point for developers, researchers, and self-hosting enthusiasts who are comfortable managing their own software. That control comes with responsibility: you maintain the instance, check its sources, and decide which services receive your data. Self-hosting is a deployment choice, not a guarantee of accuracy or privacy.
THE ESSENTIALS
More control. Fewer black boxes.
A practical foundation for your own
AI-powered research workflow.
Self-Hosted Deployment
Run your own instance with Docker, with SearXNG bundled in the standard image.
Your infrastructureLocal and Cloud LLM Support
Choose Ollama for local models or connect a supported cloud provider. Make the tradeoffs yourself.
Your choice of modelAnswers with Sources
Follow citations back to the original pages. Use sources to check claims and continue your research.
Keep the contextMAKE AN INFORMED CHOICE
Vane vs Perplexity.
More ownership or less setup?
Choose the tradeoff that fits you.
| What matters | VaneSelf-hosted & open source | PerplexityManaged AI search |
|---|---|---|
| Open-source availability | MIT-licensed source code | Hosted proprietary product |
| Self-hosting | Deploy on your own infrastructure | Use the managed service |
| Local LLM support | Connect local models through Ollama | No user-hosted local model setup |
| Setup complexity | Docker, model setup, and configuration | Open the website or app |
| Infrastructure and API costs | Your hardware, electricity, hosting, and any model API usage | Free access and paid plans; limits vary |
| Privacy considerations | Depends on your model, search services, logs, and deployment | Governed by service settings and privacy policy |
| Maintenance responsibility | You manage updates, access, and backups | Provider maintains the service |
Product details checked October 21, 2026. No benchmark or pricing claims. Sources: Vane repository and Perplexity Help Center .
Choose control or convenience. Vane puts the stack in your hands. A managed service takes infrastructure off your plate.
FROM READING TO RUNNING
Read the full installation guide Your first self-hosted search.
Get Docker ready
Install Docker Engine or Docker Desktop for your system and make sure it is running. Local LLMs need additional hardware resources.
Start the Vane container
Run the verified upstream command. The standard image bundles SearXNG; the named volume preserves application data.
Open it. Choose your model.
Visit localhost:3000 on the Docker host and complete the setup. Enter provider credentials only in your own instance.
docker run -d -p 3000:3000 -v vane-data:/home/vane/data --name vane itzcrazykns1337/vane:latestFrom the official README · checked October 21, 2026. The latest image can change; review upstream instructions before running.
Keep your instance protected.
The upstream command publishes port 3000 on all host interfaces. For local-only access, replace -p 3000:3000 with -p 127.0.0.1:3000:3000. Do not expose an unauthenticated instance to the public internet. Add access controls before remote use.
This guide never asks for or collects API keys.
A FEW THINGS TO KNOWGood questions.
Good questions.
Straight answers.
Before you build your own
search setup, start here.
What is Vane AI?
Vane is an open-source AI answering engine that combines web search with a language model to produce answers with source citations. This website is an independent guide to the project.
Is Vane the same as Perplexica?
Yes. Vane is the renamed continuation of Perplexica. Older guides may use the former name; follow the current Vane repository for installation instructions.
Is Vane a free alternative to Perplexity?
The software is available under the MIT license. Running it can still involve hardware, electricity, hosting, and cloud model API costs. Free software does not mean every deployment is free to operate.
Can I run Vane locally?
Yes. You can run a Docker instance on your own computer and access it at localhost:3000. Web search still needs internet access, even if the language model runs locally.
Does Vane support Ollama?
Yes. Ollama is supported for local models. Your Vane container must be able to reach the Ollama server; consult the official troubleshooting notes for your operating system.
Do I need an API key?
A supported cloud model provider generally requires its own credentials. Ollama can run local models without a paid cloud model API key. Configure providers in your own Vane instance, never on this guide.
Is self-hosting Vane private?
Self-hosting gives you more control, but privacy depends on configuration. Search requests reach search services, cloud models receive submitted context, and your server may retain data. Review every service involved.
Is Vane better than Perplexity?
That depends on your priorities. Vane suits people who value source access, model flexibility, and control over deployment. Perplexity may suit people who prefer a managed service with less setup and maintenance. There is no universal winner.
YOUR NEXT STEP
Get Started Make search your own.
Start with the source. Build an instance that fits your workflow.