Vane: self-hosted web answers, with sources

What Vane does well

Vane searches through SearXNG, reads sources, and uses your chosen model to compose an answer. The standard container bundles the search engine, reducing the number of services you manage. Local history and file workflows make it more than a search box.

• Choose supported local or hosted model providers instead of relying on a single subscription.
• Inspect linked sources rather than treating an answer as ground truth.
• Keep application data in your own persistent Docker volume.
• Use the slim image only when you already operate SearXNG.

Hardware and models

Vane is an application, not a model. Budget separately for the UI, SearXNG, embedding and your language model. Our 4 GB application allowance is an editorial planning estimate, not an official minimum. A remote API needs no local chat-model GPU; a local 7B model needs additional memory and a compatible inference server. Use the hardware checker before downloading weights. Live search still sends requests to external search services even when inference stays local.

A five-minute setup path

Have Docker Compose installed and a model endpoint or API key ready. Image downloads and local model downloads are additional time. Save this as compose.yml:

services:
  vane:
    image: itzcrazykns1337/vane:latest
    ports:
      - "127.0.0.1:3000:3000"
    volumes:
      - vane-data:/home/vane/data
    extra_hosts:
      - "host.docker.internal:host-gateway"
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "node", "-e", "fetch('http://127.0.0.1:3000').then(r=>process.exit(r.ok?0:1)).catch(()=>process.exit(1))"]
      interval: 30s
      timeout: 10s
      start_period: 60s
      retries: 5
volumes:
  vane-data:

docker compose config --quiet
docker compose up -d
docker compose ps

Open http://localhost:3000, complete the setup screen, select a chat model and embedding provider, then ask a short question. Follow a citation to check the result. Download this Compose file. The image tag tracks upstream changes; this guide was checked against source version 1.12.2. For repeatable installations, record the pulled image digest before upgrading.

Strengths and trade-offs

Choose it for: a direct, self-hosted search-and-answer experience, a bundled search engine and provider choice.

Plan around: model-dependent quality, search-engine availability, your own upgrades and backups. A successful health check confirms an HTTP response, not working retrieval or accurate answers. Do not assume enterprise identity or team isolation from a single-user trial.

How it differs from Perplexity

Perplexity operates a hosted service. With Vane, you choose infrastructure and providers and take responsibility for the search pipeline. You gain configuration control, but you also troubleshoot rate limits, model failures and search quality. See Vane vs Perplexity for the decision.