opencode-plugboard

Appearance

Khoj ​

Khoj is a personal AI that chats over your notes, documents and the web. One chat model writes the answers and handles intent detection and web search.

Upstream: source, docs.

Plugboard options ​

Give Khoj its own key. Docker needs a reachable address (Serving other machines).

jsonc
{
  "options": {
    "listen": { "host": "0.0.0.0", "port": 4010 },
    "keys": { "khoj": "{env:PLUGBOARD_KEY_KHOJ}" },
    "models": ["opencode-go/glm-5.3-flash"],
    "aliases": { "chat": "opencode-go/glm-5.3-flash" },
    "apps": { "khoj": { "allowedModels": ["chat"] } }
  }
}

Connect ​

Set the provider in Khoj's admin panel:

  1. Create an AI model API. Set Api Base Url to http://host.docker.internal:4010/v1 and the API key to the khoj key.
  2. Create a Chat Model. Set Model Type to Openai and the model name to chat. Leave the tokenizer unset.
  3. Open your settings and select that chat model.

The admin panel needs KHOJ_ADMIN_PASSWORD and KHOJ_DJANGO_SECRET_KEY set in the compose file. Behind a reverse proxy, also set KHOJ_ALLOWED_DOMAIN.

What works ​

Chat, intent detection and web search, all through the chat alias.

Limits ​

  • Plugboard serves no embeddings. If your Khoj search model needs a separate embedding model, run one yourself.