Vane: Self-Hosted AI Search Engine with Privacy-First Architecture

TL;DR: Vane (formerly Perplexica) is a self-hosted AI search engine that runs on your hardware. It combines SearxNG (private metasearch) with your choice of LLM — local via Ollama or cloud via OpenAI/Claude. Get Perplexity-style cited answers without your queries leaving your network. Install: docker run -d -p 3000:3000 itzcrazykns1337/vane:latest

You want AI-powered search. You don’t want your queries going to someone else’s server.

Vane gives you both.

What Is Vane?

Vane is a privacy-focused AI answering engine that runs entirely on your own hardware. It combines:

• SearxNG — Open-source metasearch engine (70+ sources)
• Your choice of LLM — Ollama (local), OpenAI, Claude, Groq, Gemini
• Cited answers — Every response includes sources

Think Perplexity, but self-hosted. Your searches never leave your network.

Note: Vane was previously known as Perplexica. Same project, same author — just rebranded.

How Do I Install Vane?

One Docker command:

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

Open http://localhost:3000 and configure your settings (API keys, models, etc.) in the setup screen.

That’s it. The image includes both Vane and SearxNG — no additional setup required.

Using Your Own SearxNG Instance

Already running SearxNG? Use the slim version:

docker run -d -p 3000:3000 \
  -e SEARXNG_API_URL=http://your-searxng-url:8080 \
  -v vane-data:/home/vane/data \
  --name vane itzcrazykns1337/vane:slim-latest

What Search Modes Does Vane Offer?

Mode	Use Case	Speed
Speed Mode	Quick lookups, simple questions	Fastest
Balanced Mode	Everyday searches	Medium
Quality Mode	Deep research, complex topics	Thorough

Choose based on your needs — trade latency for depth.

What LLM Providers Does Vane Support?

Vane is model-agnostic:

Local (fully private):

• Ollama (llama3, mistral, qwen, phi, etc.)
• Any OpenAI-compatible local server

Cloud providers:

• OpenAI (GPT-4, GPT-4o)
• Anthropic (Claude 3.5 Sonnet, Opus)
• Google (Gemini)
• Groq (fast inference)

Mix and match: Use Ollama for most queries, fall back to GPT-4 for complex research.

What Features Does Vane Include?

Feature	Description
Web search	SearxNG-powered, 70+ engines
Image/video search	Visual content alongside text
File uploads	Ask questions about PDFs, docs, images
Domain-specific search	Limit to specific websites
Smart suggestions	Query autocomplete
Widgets	Weather, calculations, stock prices
Search history	Saved locally, never lost
Discover feed	Trending content without searching

How Does Vane Compare to Alternatives?

Feature	Vane	Perplexity	Kagi	SearxNG (alone)
Self-hosted	✅	❌	❌	✅
AI answers	✅	✅	✅	❌
Local LLM support	✅	❌	❌	N/A
Cited sources	✅	✅	✅	N/A
File uploads	✅	✅	❌	❌
Free	✅	Freemium	Paid	✅
Privacy	Full	Limited	Good	Full

Vane’s unique value: Self-hosted AI search with local LLM support. No cloud required.

How Do I Connect Ollama?

For fully private, local-only search:

1. Run Ollama on your machine
2. In Vane settings, set the Ollama API URL:
   • Windows/Mac: http://host.docker.internal:11434
   • Linux: http://<your-ip>:11434
3. Select your model (llama3.2, mistral, etc.)

Linux users: Expose Ollama to the network:

# Edit /etc/systemd/system/ollama.service
# Add: Environment="OLLAMA_HOST=0.0.0.0:11434"
sudo systemctl daemon-reload
sudo systemctl restart ollama

Can I Use Vane as My Default Search Engine?

Yes. Add this to your browser’s custom search engines:

http://localhost:3000/?q=%s

Replace localhost:3000 with your actual Vane URL if hosted elsewhere.

What’s the Architecture?

Vane is built on Next.js with a modular architecture:

1. Frontend — Clean search interface with mode selection
2. Search layer — SearxNG for web results (Tavily/Exa support coming)
3. LLM layer — Your choice of provider for answer generation
4. Widget system — Contextual cards (weather, math, stocks)
5. Storage — Local SQLite for history and settings

All components run in a single container. Data persists in the mounted volume.

What’s Coming Next?

The roadmap includes:

• More widgets and integrations
• Additional search sources (Tavily, Exa)
• Custom agents
• Authentication support

When Should I Use Vane?

Use Vane when:

• You want Perplexity-style search but self-hosted
• Privacy is non-negotiable
• You’re already running Ollama
• You want to mix local and cloud models

Consider alternatives when:

• You need mobile apps (Perplexity)
• You want managed infrastructure (Kagi)
• You don’t need AI answers (plain SearxNG)

Frequently Asked Questions

What is Vane?

Vane is a privacy-focused AI answering engine that runs on your own hardware. It combines SearxNG metasearch with local or cloud LLMs to deliver cited answers without your data leaving your network.

How do I install Vane?

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

Does Vane work completely offline?

With Ollama and cached search results, Vane can work largely offline. However, web search requires internet connectivity to fetch fresh results.

Is Vane a fork of Perplexica?

Vane shares some DNA with the Perplexica project but has evolved with its own feature set, including widgets, file uploads, and a different architecture.

How much does Vane cost?

Vane is free and open source. You pay only for cloud LLM API calls if you choose to use them. Local Ollama usage is completely free.

Can I run Vane on a Raspberry Pi?

Theoretically yes, but performance depends on your LLM. SearxNG runs fine on Pi hardware. For LLM inference, consider using cloud providers or a more powerful machine.

What’s the difference between full and slim Docker images?

The full image (latest) includes bundled SearxNG. The slim image (slim-latest) is smaller and connects to your existing SearxNG instance.

Links:

• GitHub Repository
• Architecture Documentation
• Discord Community
• Docker Hub