Vane: Your Own Private AI Answer Engine That Runs on Your Hardware

Ever wish you could get AI search results without sending your queries to some cloud server? That's exactly what Vane does. It's a privacy-focused AI answer engine that runs entirely on your own hardware. No data leaves your machine. No API keys. No subscription fees.

It's basically Perplexity AI but self-hosted and completely offline. Built on open source tech, Vane gives you AI-powered answers from your own indexed content or public sources, all while keeping your searches private.

What It Does

Vane is a self-hosted answer engine. You feed it documents, websites, or your local files (via a vector database), and it uses a local LLM to answer questions based on that content. Think of it as a private research assistant that only knows what you tell it.

Under the hood, it uses:

• A vector database (like Chroma or Qdrant) to store embeddings of your documents
• A local LLM (via Ollama or compatible endpoints) for generating answers
• A search component that retrieves relevant chunks from your indexed data

You ask a question in natural language. It searches your indexed content, retrieves relevant passages, and generates a coherent answer with citations. All done locally.

Why It's Cool

Complete privacy. Your queries, your data, your hardware. No third party sees anything. This is huge for developers who work with sensitive code, proprietary documentation, or personal knowledge bases.

Customizable knowledge. You control what the AI "knows." Index your company's internal docs, your personal notes, or a specific set of research papers. The model only answers from your data.

No API costs. Once it's running, it's free. No per-query charges. No rate limits. Just your electricity bill.

Developer friendly. It's a CLI tool with clean JSON APIs. You can script it, integrate it into your workflows, or wrap it in a web UI. The GitHub repo has decent documentation for getting started.

Offline capable. Once you download the models, you can run it without internet access. Great for air-gapped environments or when you're coding on a plane.

How to Try It

Clone the repo and follow the README:

git clone https://github.com/ItzCrazyKns/Perplexica
cd Perplexica

You'll need:

• Python 3.10+
• A running instance of Ollama (for local LLM)
• A vector database (Chroma is easiest to get started)

Install dependencies:

pip install -r requirements.txt

Configure your settings in config.yaml (point it to your Ollama endpoint, set the model, define your data sources).

Ingest your data:

python ingest.py --source /path/to/your/docs

Then run the CLI:

python query.py "What does the authentication module do?"

Or use the API server:

python server.py

Then hit http://localhost:8000/ask with your questions.

The README has Docker instructions too if that's your jam.

Final Thoughts

Vane isn't trying to replace ChatGPT or Claude. It's for a specific use case: private, local, customizable question answering. If you've ever wanted an AI assistant that only knows your stuff and doesn't phone home, this is it.

I could see this being really useful for:

• Teams that need internal documentation search without data leaving the network
• Developers who want to query their own codebases or notes
• Privacy-conscious users who want AI answers without the tradeoff

It's rough around the edges (it's a young project), but the core idea is solid. Try it out, contribute if you can, and keep your queries to yourself.

Follow @githubprojects for more open source tools like this.