LLM as a Service (LLMaaS)

LLMaaS provides access to AI models hosted by Cloud Temple via an API: conversational assistants, coding and agents, image analysis, semantic search, reranking, translation, and audio/image use cases.

Catalog, Updates, and Availability

• Model Catalog and Lifecycle : models, context, LTS statuses, end-of-support dates, and recommended migrations.
• LLMaaS Updates : model additions, service updates, and redirection announcements.
• Service Status and History : observed availability and performance.

To get started, follow the quickstart guide. To choose a model, refer to the catalog guide.

Features

Usage	Guide
Text generation, streaming, tool calls, and vision	API LLMaaS
Retrieval-Augmented Generation	Understanding RAG
Document relevance ranking	Reranking
Asynchronous batch request processing	Batch
Model switching and legacy ID management	Lifecycle and migration

Inference optimizations, including the multi-token prediction (MTP) announced in the changelog, are handled on the service side. Evaluate their impact on your own requests using monitoring metrics.

API Access

The API is accessible via the Cloud Temple Console. You can manage your API keys, monitor your usage, and configure your tiers in your account settings. The console also allows you to visualize your model usage.

Authentication

All requests to the LLMaaS API must include an Authorization header with your API key in Bearer token format. If you are using the client SDKs, the key will be automatically included in each request. If you integrate directly with the API, you must send this header yourself.

Content Types

Text generation requests use JSON with the header Content-Type: application/json. Streaming returns SSE events and audio file transcription uses multipart/form-data. Refer to the API documentation for the format of each endpoint.

Response Metadata

Depending on the endpoint, the JSON response body may contain the following fields:

• id : Response identifier.
• backend : Information about the engine and instance that processed the request (engine_type, machine_name).

These fields are read from the JSON body, not from the HTTP headers. Consult the API reference for the format and documented fields of each endpoint.

Examples

cURL Request

curl -X POST "https://api.ai.cloud-temple.com/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "gpt-oss:120b",
    "messages": [
      {
        "role": "user", 
        "content": "Salut ! Peux-tu te présenter en français ?"
      }
    ],
    "max_tokens": 200,
    "temperature": 0.7
  }'

Response

{
  "backend": {
    "engine_type": "engo",
    "machine_name": "ma02"
  },
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "content": "Bonjour ! Je suis un modèle de langage virtuel...",
        "role": "assistant"
      }
    }
  ],
  "created": 1749110753,
  "id": "chatcmpl-ollama-14b812ef-b21f-430c-b93c-d0d1bf653806",
  "model": "gpt-oss:120b",
  "object": "chat.completion",
  "usage": {
    "completion_tokens": 200,
    "prompt_tokens": 70,
    "reasoning_tokens": 0,
    "total_tokens": 270
  }
}

Available Parameters

Parameter	Type	Description
model	string	The model to use (see catalogue des modèles)
messages	array	List of conversation messages
max_tokens	integer	Maximum number of tokens to generate
temperature	float	Controls creativity (0.0-2.0)
top_p	float	Controls response diversity
stream	boolean	Enables response streaming
user	string	Unique identifier for the end user

Base URL

The base URL for all API requests is:

https://api.ai.cloud-temple.com/v1/

Available Endpoints

• /chat/completions : Generation of conversational responses
• /chat/completions/batch : Asynchronous processing of multiple conversations (guide Batch)
• /completions : Simple text completion
• /embeddings : Vectorization for semantic search and RAG
• /rerank and /v2/rerank : Result reranking (compatible Cohere SDK)
• /audio/transcriptions : Batch audio transcription (Whisper)
• /audio/speech : Speech synthesis (TTS)
• /images/generations : Image generation
• /models : List of available models

Example: List of Models

curl -X GET "https://api.ai.cloud-temple.com/v1/models" \
  -H "Authorization: Bearer YOUR_API_KEY"

The response provides the identifiers exposed by the API. Also consult the lifecycle to check for deprecations and redirects. The models used in the examples are for illustrative purposes; verify their availability before execution.