Product explainer
What is Vane AI?
Vane is an open-source AI answering engine that you run yourself. It is one self-hosted alternative for a Perplexity-like web-answer workflow. Formerly called Perplexica, it combines web results from SearXNG with a model you choose to generate answers that cite sources. Its official project is maintained at ItzCrazyKns/Vane.
Project description and setup options checked October 22, 2026 against the official repository. This page explains the published architecture; it is not a hands-on performance review.
How an answer is made
The official answer-flow guide describes classification, optional research, answer generation, and citation display. In a web-research question, Vane can retrieve results through its configured search backend before the selected model drafts an answer. You should open the cited pages to check each claim. Merely opening the setup screen does not prove search and generation work together.
Where a web question can travel
- Your browser → VaneThe interface and application run on hardware you control.
- Vane → search layer → web sourcesWhen web research is enabled, the SearXNG search layer can contact outside search services. A local model does not make live web lookup offline.
- Vane → selected modelOllama can keep inference on your hardware. A selected cloud provider receives the request data sent to it.
- Answer → your browserVane displays the response and source links; you still need to verify whether each link supports its claim.
These boundaries follow the answer-flow guide and README's search/model options. Exact data handling depends on your version and chosen providers.
Models and deployment
Vane's documentation lists local-model support through Ollama and cloud-provider options. Its recommended Docker image bundles SearXNG; the documented one-container command is in our installation guide. If Ollama runs on the Docker host, the address visible from inside the container differs from the host's own localhost; see our connection guide.
Is Vane free?
The official package identifies the project as MIT-licensed. Running it can still use paid hosting or a paid cloud model, and a local model consumes computer resources. See the cost breakdown before describing a particular setup as free.
Who is it for?
Vane fits a person who wants a Perplexity-like web-answer workflow and is willing to run and maintain the service. It is less convenient if you need a hosted product with no setup. If you mainly want a chat workspace, a document assistant, a research agent, or only raw search results, the alternatives comparison separates those tool types.
What are the main limitations of Vane?
- Answer quality depends on retrieved sources and the selected model; a citation is a link to inspect, not automatic proof.
- Self-hosting does not make external web search offline.
- Using a cloud model can send data to that provider.
- You are responsible for updates, availability, and access control on your own installation.
Does Vane have an API?
Yes. The project's Search API documentation describes provider discovery and a programmatic search endpoint. Check that documentation for the request fields and your installed version before integrating it. This guide has not tested API calls against a running Vane instance.
Is Perplexica a separate product?
No. The project's README identifies Vane as formerly Perplexica. The name-change guide explains what to check when an older tutorial still says Perplexica. This independent site is neither the official Vane project nor a Vane hosting service.
Source
Official Vane repository and README and Search API documentation, checked October 22, 2026.