Chat Completions API

POST /v1/chat/completions

The /v1/chat/completions endpoint is the central entry point for conversational and text-generation applications. It supports streaming, tool calling, and multimodal input (vision).

Request Parameters

Parameter	Type	Description
model	string	Model ID, e.g. vllm/release/gpt-oss-120b
messages	array	List of messages (role-driven)
temperature	float	Sampling temperature (default: 1.0)
top_p	float	Nucleus sampling (default: 1.0)
max_tokens	integer	Maximum number of generated tokens
stream	boolean	Enables SSE streaming
tools	array	Available tools (function definitions)
tool_choice	string/object	Controls tool selection

Message Format

Messages follow a role model with three standard roles:

• system: controls the model’s behavior and persona
• user: user input
• assistant: previous model responses (for multi-turn)

{
  "model": "vllm/release/gpt-oss-120b",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "What is the capital of France?"},
    {"role": "assistant", "content": "Paris."},
    {"role": "user", "content": "How many inhabitants does the city have?"}
  ]
}

Basic Example

curl

curl -X POST https://ai.noris.de/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "vllm/release/gpt-oss-120b",
    "messages": [{"role": "user", "content": "Explain RAG in one sentence."}]
  }'

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://ai.noris.de/v1",
    api_key="YOUR_API_KEY"
)

response = client.chat.completions.create(
    model="vllm/release/gpt-oss-120b",
    messages=[{"role": "user", "content": "Explain RAG in one sentence."}]
)
print(response.choices[0].message.content)

JavaScript

const response = await fetch("https://ai.noris.de/v1/chat/completions", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "Authorization": "Bearer YOUR_API_KEY"
  },
  body: JSON.stringify({
    model: "vllm/release/gpt-oss-120b",
    messages: [{role: "user", content: "Explain RAG in one sentence."}]
  })
});
const data = await response.json();
console.log(data.choices[0].message.content);

Streaming

Set "stream": true to receive responses token by token as Server-Sent Events (SSE). This reduces perceived latency, since the first tokens are emitted immediately.

See Streaming for details and examples.

Tool Calling

The tools array exposes functions to the model. Based on tool_choice, the model decides whether to invoke a tool and returns the call in the response as tool_calls.

See Tool Calling for details and examples.

Multimodal/Vision

For multimodal models, content is passed as an array that can also include image_url entries:

{
  "role": "user",
  "content": [
    {"type": "text", "text": "What does this image show?"},
    {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}}
  ]
}

See Multimodal for details and examples.

Response Structure

A successful response contains a choices array. Each choice includes the generated message plus finish-reason metadata. The usage block reports the tokens processed:

Field	Meaning
choices[].message.content	Generated text
choices[].finish_reason	Reason for termination (e.g. stop, length)
usage.prompt_tokens	Input tokens
usage.completion_tokens	Generated tokens
usage.total_tokens	Sum of input and output

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "choices": [
    {
      "index": 0,
      "message": {"role": "assistant", "content": "..."},
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 28,
    "total_tokens": 40
  }
}