API Reference

The Xinference REST API is built with FastAPI and follows OpenAPI conventions. Routes live under two prefixes: management resources (deployments, deployable models) are under /api/v1, while account, organization, billing, and the OpenAI-compatible inference proxy are under /v1.

Base URL

https://api.xinference.co

Authentication

The management API authenticates with an HttpOnly session cookie issued when a user signs in (email/password) or completes Google SSO; these endpoints reject Authorization headers. The inference endpoints (/v1/chat/completions, /v1/embeddings) additionally accept an API key as Authorization: Bearer xi-sk-…. See Authentication for sign-in, SSO, password reset, and API keys.

Content Type

All request and response bodies use JSON:

Content-Type: application/json

Pagination

List endpoints that paginate accept limit and offset query parameters. For example, the deployments list:

GET /api/v1/model-deployments?limit=100&offset=0

Paginated responses include:

{
  "items": [ ... ],
  "total": 142,
  "limit": 100,
  "offset": 0
}

Health Check

GET /healthz

Returns {"status": "ok"} when the API is healthy. No authentication required.

API Sections

• Deployments API →
• Models API →
• Organizations API →
• Billing API →

OpenAPI Spec

The interactive OpenAPI documentation is served by FastAPI at:

GET /docs        # Swagger UI
GET /redoc       # ReDoc UI
GET /openapi.json

Inference & SDK Compatibility

The inference endpoints (/v1/chat/completions and /v1/embeddings) are OpenAI-compatible: they accept OpenAI-style request bodies and return OpenAI-style responses, proxied to your running deployment. They authenticate with an API key (Authorization: Bearer xi-sk-…) or the session cookie, so the official OpenAI SDKs work by pointing the base URL at Xinference. See the Inference API for details.