Connect a client

llama-dash exposes one OpenAI- and Anthropic-compatible /v1/* endpoint. Point any client that supports a custom base URL at llama-dash instead of the inference backend.

Base URL

Use the llama-dash origin with a /v1 suffix:

http://localhost:3000/v1

Authentication

If any API keys exist in the database, the proxy requires a bearer token:

Authorization: Bearer sk-...

Create a key from API Keys before connecting a real client. The secret is shown once. If no keys exist yet, the proxy accepts unauthenticated requests, so you can test a new install without an Authorization header.

First request

curl http://localhost:3000/v1/chat/completions \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "your-model-id",
    "messages": [{ "role": "user", "content": "Hello!" }]
  }'

Replace your-model-id with the ID from the Models page. The example includes an API key because that is the normal operating mode. Remove the Authorization header only when your installation has no keys yet.

The request appears in the Requests log within the dashboard, with routing, token counts, latency, and an estimated cost.

More clients

The dashboard's Endpoints page generates ready-to-paste snippets for many clients. See Connecting clients for curl, Python, TypeScript, Open WebUI, Home Assistant, opencode, and Claude Code.