Raiana Models API

Authentication

These endpoints do not currently enforce an Authorization header, unlike /v1/chat/completions and /v1/responses. We recommend always sending your API key anyway, since this may become required in the future:

Authorization: Bearer raikey_YOUR_API_KEY_HERE

List Models

GET /v1/models

Returns a list of all available models.

Example Request

curl https://api.chatmdr.eu/v1/models \
  -H "Authorization: Bearer raikey_YOUR_API_KEY_HERE"

Response Format

{
  "object": "list",
  "data": [
    {
      "id": "chatmdr-fast-openai",
      "object": "model",
      "owned_by": "ChatMDR",
      "created": 1751373000
    },
    {
      "id": "chatmdr-smart-openai",
      "object": "model",
      "owned_by": "ChatMDR",
      "created": 1751373000
    }
  ]
}

Response Fields

Error Handling

Retrieve Model

GET /v1/models/{model_id}

Returns details about a specific model.

Path Parameters

Example Request

curl https://api.chatmdr.eu/v1/models/chatmdr-fast-openai \
  -H "Authorization: Bearer raikey_YOUR_API_KEY_HERE"

Response Format

{
  "id": "chatmdr-fast-openai",
  "object": "model",
  "owned_by": "ChatMDR",
  "created": 1751373000
}

Response Fields

Same shape as one entry of data[] in List Models above: id, object, owned_by, created.

Error Handling

Available Models

Each model ID identifies both a product (which regulatory domain it answers questions about) and a tier (response speed/depth trade-off). Fast models optimize for low latency and shorter answers; Smart models allow longer, more thorough answers with deeper reasoning. All models support web search and file/document grounding when building an answer.

ChatMDR — EU Medical Device Regulation

ChatIVDR — EU In Vitro Diagnostic Regulation

ChatFDA — US FDA Medical Device Regulation

ChatAIAct — EU AI Act

* Default output cap. Callers can override it per request (max_tokens on /chat/completions, max_output_tokens on /responses), clamped to the range 100..4096.

Raiana Chat Completions API

Authentication

All requests must include an API key in the Authorization header using the Bearer token scheme:

Authorization: Bearer raikey_YOUR_API_KEY_HERE

Raiana API keys use the raikey_ prefix.

Create Chat Completion

POST /v1/chat/completions

Given a list of messages, returns a single model-generated reply. This endpoint is not streamed — for a streamed alternative, see the Responses API.

Request Body Parameters

Not supported on this endpoint: stream, tools / tool_choice / functions / function_call, multiple choices (n), and multimodal message content — content must be a plain string.

Message Roles

Example Request

curl -X POST https://api.chatmdr.eu/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer raikey_YOUR_API_KEY_HERE" \
  -d '{
    "model": "chatmdr-fast-openai",
    "messages": [
      {
        "role": "user",
        "content": "What is a Class III medical device under MDR?"
      }
    ],
    "max_tokens": 1024
  }'

Response Format

{
  "object": "chat.completion",
  "id": "825c3790-ec66-4008-8b30-99efd968707c",
  "model": "chatmdr-fast-openai",
  "created": 1751373000,
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Under MDR, Class III devices are the highest risk class..."
      }
    }
  ],
  "usage": {
    "input_tokens": 5820,
    "output_tokens": 387,
    "total_tokens": 6207,
    "input_tokens_details": { "cached_tokens": 0 },
    "output_tokens_details": { "reasoning_tokens": 0 }
  }
}

Response Fields

Error Handling

Raiana Responses API (beta)

Do not use this API for production use yet. You can report bugs to info@raiana.ai.

Authentication

All requests must include an API key in the Authorization header using the Bearer token scheme:

Authorization: Bearer raikey_YOUR_API_KEY_HERE

Create Response

POST /v1/responses

Given text input, returns a single model-generated reply — either as a complete JSON object, or streamed as Server-Sent Events. This is a text-only, OpenAI Responses-compatible endpoint.

Request Body Parameters

Not supported on this endpoint: tools, tool_choice, functions, function_call, and parallel_tool_calls (all rejected with 400), non-text modalities, and non-text content parts inside input (images/audio/video).

Input Items

When input is an array, each item is either a plain string (treated as a user message) or a message object:

Each content part object has a type of input_text, text, or output_text, and a text string field.

Example Request

curl -X POST https://api.chatmdr.eu/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer raikey_YOUR_API_KEY_HERE" \
  -d '{
    "model": "chatmdr-smart-openai",
    "input": "What are the MDR classification rules for active implantable devices?",
    "max_output_tokens": 1024
  }'

Response Format

{
  "id": "resp_825c3790",
  "object": "response",
  "created_at": 1751373000,
  "status": "completed",
  "model": "chatmdr-smart-openai",
  "output": [
    {
      "id": "msg_resp_825c3790",
      "type": "message",
      "status": "completed",
      "role": "assistant",
      "content": [
        {
          "type": "output_text",
          "text": "Active implantable devices are classified as Class III...",
          "annotations": []
        }
      ]
    }
  ],
  "output_text": "Active implantable devices are classified as Class III...",
  "usage": {
    "input_tokens": 5820,
    "output_tokens": 387,
    "total_tokens": 6207,
    "input_tokens_details": { "cached_tokens": 0 },
    "output_tokens_details": { "reasoning_tokens": 0 }
  },
  "conversationId": "resp_825c3790"
}

Response Fields

Streaming

Pass "stream": true to receive the response as Server-Sent Events instead of a single JSON object.

curl -X POST https://api.chatmdr.eu/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer raikey_YOUR_API_KEY_HERE" \
  -d '{
    "model": "chatmdr-smart-openai",
    "input": "Summarize Article 10 obligations for manufacturers.",
    "stream": true
  }'

The response has Content-Type: text/event-stream and emits a sequence of named events, each with a data: payload, ending in a final data: [DONE] line:

Example Event

event: response.output_text.delta
data: {"type":"response.output_text.delta","delta":"Active ","output_index":0,"content_index":0}

Error Handling