Vane (formerly Perplexica)

Vane is a self-hosted answering engine. It searches the web through SearXNG and answers with cited sources, using an LLM for the answer and an embedding model for ranking.

Upstream: source.

Plugboard options

Plugboard serves the answer model only. Vane's chat and search requests also carry an embedding model, and plugboard has no /v1/embeddings. Pick a second provider for embeddings before you install.

Give Vane its own key and a chat alias. Docker needs a reachable address (Serving other machines).

{
  "options": {
    "listen": { "host": "0.0.0.0", "port": 4010 },
    "keys": { "vane": "{env:PLUGBOARD_KEY_VANE}" },
    "models": ["anthropic/claude-sonnet-4-6"],
    "aliases": { "answer": "anthropic/claude-sonnet-4-6" },
    "apps": { "vane": { "allowedModels": ["answer"] } }
  }
}

Connect

Vane sets providers in its setup screen. The OpenAI provider has two fields:

Field	Value
API Key	the vane key (env var OPENAI_API_KEY)
Base URL	http://host.docker.internal:4010/v1, ending in /v1 (env var OPENAI_BASE_URL)

Then set:

1. Chat model: the alias answer.
2. Embedding model: a model from your other provider, not a plugboard alias.

Run Vane with Docker:

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

What works

Answers through the answer alias, once the embedding model is set.

Limits

• Chat, search and uploads all need the embedding model from the second provider.
• Vane has no login. Keep its port on loopback or a trusted network.
• Enter the plugboard key in the setup screen or OPENAI_API_KEY. Do not bake it into an image or commit it.