API Endpoints

AI Foundation Services provides an OpenAI-compatible REST API. All endpoints use the base URL:

https://llm-server.llmhub.t-systems.net/v2

For the complete OpenAPI specification, see the interactive API docs (Redoc).

Endpoints

Method	Endpoint	Description
GET	/models	List all available models
GET	/models/{model_id}	Get model details and metadata
POST	/chat/completions	Create a chat completion
POST	/completions	Create a text completion
POST	/embeddings	Create embeddings
POST	/audio/transcriptions	Transcribe audio to text
POST	/audio/translations	Translate audio to English
GET	/audio/models	List available audio models
POST	/images/generations	Generate images from text
POST	/responses	Create a response (Responses API)
POST	/files	Upload a file
GET	/files	List uploaded files
DELETE	/files/{id}	Delete a file

Queue Endpoints

All major endpoints have /queue variants. They work like /v2, but when you are over your rate limit they wait until it frees up instead of returning 429 right away. They are not asynchronous. See the Queued Requests guide.

Method	Endpoint	Description
POST	/queue/chat/completions	Chat completion (queued)
POST	/queue/completions	Text completion (queued)
POST	/queue/embeddings	Embeddings (queued)
POST	/queue/audio/transcriptions	Audio transcription (queued)
POST	/queue/audio/translations	Audio translation (queued)
POST	/queue/images/generations	Image generation (queued)
POST	/queue/images/edits	Image editing (queued)
GET	/queue/models	List models (queue)

API Versioning

The API uses versioned paths:

Path Prefix	Purpose	Examples
/v2	Default — LLM inference, embeddings, audio, images	/v2/chat/completions, /v2/embeddings
/v1	Visual RAG, vector stores, file management	/v1/vector_stores, /v1/files
/queue	Same as /v2, but waits up to two minutes when over the rate limit	/queue/chat/completions

• Use /v2 for all standard LLM API calls (this is the default when using OpenAI SDKs with our base URL)
• Use /v1 for Visual RAG and file management operations
• Use /queue for batch jobs that sometimes go over the rate limit

Authentication

All requests require an API key in the Authorization header:

Authorization: Bearer YOUR_API_KEY

See Authentication for setup details.

Request Format

All POST requests use JSON:

curl -X POST "https://llm-server.llmhub.t-systems.net/v2/chat/completions" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-oss-120b", "messages": [{"role": "user", "content": "Hello"}]}'

Response Format

Responses follow the OpenAI response format. For chat completions:

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1710000000,
  "model": "gpt-oss-120b",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello! How can I help you today?"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 10,
    "completion_tokens": 9,
    "total_tokens": 19
  }
}