Does Vane require a domain name?

No. Vane works on a plain IP address. Access it via SSH tunnel (ssh -L 3000:127.0.0.1:3000 root@ ) then open http://localhost:3000 in your browser, or configure a custom domain with ServOrbit for browser-direct access.

Can Vane work without any external LLM API key?

Yes. Vane can run in web-search-only mode (SearXNG results, no AI synthesis) without any API key. For AI answers, you can use a local Ollama model on the same VPS instead of a paid API — no data leaves your server.

What is the relationship between Vane and Perplexica?

Vane is the official rebranding of Perplexica (March 2026). The codebase, features and Docker image moved to a new repository under the same author (ItzCrazyKns). The bundled SearXNG, MIT license and single-container approach are unchanged.

Does Vane require an external database?

No. Vane stores all data (conversation history, settings, config) in SQLite inside the /home/vane/data volume. No PostgreSQL or MySQL instance is needed. Back up the volume to preserve your data.

How does first access work?

Vane is open by default — there is no login page. The interface is accessible immediately after deployment. LLM providers are configured in Settings (top-right icon) after first launch, but the app is usable without them (web search only).

Build, host and run AI solutions.

Vane

Vane — self-hosted AI search engine: SearXNG + LLM in one container, no cloud subscription.

1 GB minimum RAM 1 vCPU Port 3000 In validation

In validation

Automatic installation for this solution is ready and currently going through our tests on a real server. Ordering will open as soon as validation is complete.

Source code

Tech stack

DockerNext.jsSearXNGSQLite

Minimum RAM1 GB minimum

Minimum CPU1 vCPU

Default port3000

Compatible OSChoice of Linux distributions

Vane (formerly Perplexica) is an open-source AI search engine (MIT) that combines SearXNG's privacy-first metasearch with the reasoning power of your own LLM. Deploy a single Docker container on your VPS and get a Perplexity-like interface — search the web, get synthesised answers with cited sources — without any data leaving your infrastructure.

Vane bundles SearXNG internally (port 8080) and serves the Next.js chat interface on port 3000. It connects to any compatible LLM: a local Ollama instance, OpenAI, Google Gemini, Groq or Anthropic. The SQLite database stores conversation history and settings locally. No account creation is required — open the URL and start searching immediately.

Key features

Bundled SearXNG — a privacy-focused metasearch engine (Google, Bing, DuckDuckGo…) runs inside the same container; no second service to configure.

LLM-agnostic — connect Ollama (local models), OpenAI, Google Gemini, Groq or Anthropic; change provider at any time in Settings.

Cited answers — every AI-synthesised response lists its web sources with links, so you can verify claims immediately.

Conversation history — past searches are stored locally in SQLite; revisit and continue any conversation thread.

No-auth access — open the URL and search immediately; no sign-up, no login, no user management overhead.

Single container — SearXNG and the Next.js frontend ship inside one image; one volume persists all data.

When to use this solution?

1

Replace Perplexity for a development team

Give your team a private AI search with no subscription cost and no usage caps. Queries go through your own SearXNG instance; answers come from your Ollama model or an OpenAI key shared at the team level. Conversation history stays on your VPS.

2

Research with cited web sources

Ask a technical question and get an LLM-synthesised answer with direct links to the source pages. Ideal for researching libraries, comparing frameworks or monitoring technology trends without manual tab-switching.

3

Privacy-first web search

SearXNG aggregates results from multiple engines without tracking queries or sharing your IP with search providers. Vane adds AI synthesis on top while keeping every request within your own infrastructure.

Deploy Vane on your VPS

Guide optimized for ServOrbit Cloud VPS.

01

Order a ServOrbit VPS

Vane runs on a single container. A 1 GB RAM VPS with Ubuntu 24.04 is sufficient for most setups. If you plan to run a local Ollama model on the same server, provision at least 4 GB RAM and additional disk space for model weights.

02

Install Vane from the Marketplace

In the ServOrbit client area, select Vane in the AI category and click Install. The provisioning script deploys the container, maps port 3000 via the nginx reverse proxy, and starts Vane automatically.

03

Open the interface

Navigate to your VPS address (or custom domain) in a browser. Vane launches directly — no account creation required. The first search uses SearXNG without LLM synthesis until you configure a provider.

04

Connect an LLM

Click the Settings icon in the top-right corner. Under Model Providers, enter your Ollama URL (e.g. http://localhost:11434) or an API key for OpenAI, Gemini, Groq or Anthropic. Save — Vane will now return AI-synthesised answers with sources.

05

Choose a search focus mode

Vane offers focus modes to scope searches: Web, Academic, YouTube, Reddit, Writing Assistant or Wolfram Alpha. Select a mode from the input bar before submitting your query to get more targeted results.

Frequently asked questions

No. Vane works on a plain IP address. Access it via SSH tunnel (ssh -L 3000:127.0.0.1:3000 root@<ip>) then open http://localhost:3000 in your browser, or configure a custom domain with ServOrbit for browser-direct access.

Complementary solutions

Development

Vane isn't orderable yet.