OpenAPI and Postman

These assets are portable representations of the current LLMTR Gateway API behavior. They do not add an endpoint. Model-specific capabilities and pricing can change, so they are not frozen into the OpenAPI document; read them from the unauthenticated GET /v1/models response.

Downloadable assets

• OpenAPI 3.1 document
• Postman collection
• Postman environment

The api_key value in the Postman environment is intentionally empty. After you import the environment into a shared workspace, define the key as a secret only in your local Postman environment. Never put the key in the collection, an environment export, or Git history.

Read-only smoke

Requests in the collection's Public model discovery folder need neither an API key nor credit. The GET /v1/models test checks that the response is a model list and stores the first suitable Chat Completions model in the chat_model variable. Authenticated examples can then use that value.

Authenticated requests can make a real call and incur cost. Before sending one, check the selected model's supported_endpoints, pricing, and limit fields in the model-discovery response.

Security schemes

Gateway POST requests accept the same LLMTR API key in either of these forms:

Authorization: Bearer llmtr-your_key

x-api-key: llmtr-your_key

The second form supports Anthropic-compatible clients. Model-discovery GET endpoints are anonymous.

Contract scope

The OpenAPI document describes Chat Completions, Messages, Responses, image, audio, research, robotics, embeddings, rerank, video, realtime, files, and model-discovery surfaces. A model is not guaranteed to support every surface. Use the model card's supported_endpoints field before making a call.

For working TypeScript and Python examples, see the Cookbook quickstarts.