Self-Hosted / Projects / Vane
Vane
StableSpecs verified September 24, 2026 · against vlatest
Overview
Self-hosted AI answering engine for private web search, citations, and local LLMs.
Vane is a self-hosted AI answer engine for people who want a Perplexity-style research workflow under their own control. It combines search results with an LLM response and cited sources instead of returning only a conventional list of links.
Why people choose Vane
The real search intent is private AI search: ask a question, retrieve current web information, and inspect the sources used in the answer. Vane is useful for homelab operators and developers who want an open-source answering interface that can connect to local or hosted model providers.
What is included
The official repository describes an AI-powered search and answering engine with a web interface, source citations, agent-oriented features, SearXNG integration, Docker deployment files, and support for configurable model providers. Exact provider behavior depends on the version and configuration you deploy.
Requirements and tradeoffs
Vane is not a single static website: a useful deployment needs the application, a search backend such as SearXNG, an LLM provider, persistent application data, and secrets configured through the project setup. Local models can improve privacy but require suitable hardware; hosted models may be easier but send prompts or retrieved content to an external provider.
Best fit
Choose Vane when you want a self-hosted AI search and answer interface with sources. If you need a general-purpose personal knowledge base, document chat, or production search platform, compare it with tools such as Khoj, Open WebUI, or a dedicated search stack rather than treating Vane as all of them.
Interface previews
Screenshots of the Vane interface.
Feature support
| Feature | Support |
|---|---|
| AI-generated answers with citationsThe official project presents Vane as an AI-powered answering engine and search experience with source references. | Supported |
| Web search integrationThe repository includes SearXNG in the project stack and describes Vane as an AI search engine. | Supported |
| Local LLM compatibilityThe project supports configurable model providers, but the exact local-model path depends on the deployed version and provider configuration. | Partial support |
| Hosted LLM provider supportProvider availability and required environment variables should be checked against the current repository setup instructions. | Partial support |
| Docker deploymentThe official repository includes a Dockerfile and docker-compose.yaml deployment material. | Supported |
| Research-oriented web interfaceThe project is designed around asking questions and returning synthesized answers from searched sources. | Supported |
| Persistent application dataThe repository contains database and migration material, so operators should plan persistent storage and backups. | Supported |
| Privacy-preserving deploymentSelf-hosting keeps the application under your control, but prompts and retrieved data may still leave the server when a hosted model or external service is configured. | Partial support |
Community signals
- Stars
- 36.9k
- Forks
- 4.1k
- Open issues
- 355
- Last commit
- 4w ago
- Latest release
- v1.12.2
- Repo created
- Apr 2024
Includes open pull requests
6mo ago
Refreshed nightly from the GitHub API.
Questions
What is Vane used for?
Vane is used to search the web and produce AI-generated answers with cited sources through a self-hosted interface.
Is Vane a self-hosted Perplexity alternative?
Vane targets a similar AI-answer and cited-search workflow and can be self-hosted. Compare its current model, search, and deployment integrations with the hosted service you want to replace.
Can Vane use local LLMs?
Vane supports configurable model providers, and local-model use depends on the current release and provider configuration. Check the repository setup instructions before assuming a particular runtime is supported.
Does Vane need SearXNG?
The official repository includes SearXNG in the project stack and uses it as the search component in the documented deployment path, so plan for a search backend rather than only the application container.
Can you deploy Vane with Docker?
Yes. The official repository includes Docker and Docker Compose deployment files, although you still need to configure model, search, database, and secret settings.
Is Vane fully private?
Self-hosting controls where the application runs, but privacy also depends on the selected LLM, search provider, telemetry, and network configuration. Hosted providers can receive prompts or retrieved context.
Is Vane a replacement for a knowledge base?
Not by itself. Vane focuses on AI-powered web answering and search; a document library, personal knowledge base, or retrieval system may be a better fit for private notes and long-term archives.
support // the lab
Found this write-up useful?
If it saved you time or a rebuild, you can support more practical homelab guides.