> ## Documentation Index
> Fetch the complete documentation index at: https://docs.modelslab.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Chat Completions

> Send a conversation to your Enterprise LLM Endpoint and get the model's reply. OpenAI-compatible, with streaming.

## Request

Send a `POST` request with your enterprise API key as a Bearer token.

```curl curl theme={"theme":{"light":"github-light","dark":"github-dark"}}
--request POST 'https://modelslab.com/api/v1/enterprise/proxy/v1/chat/completions' \
--header 'Authorization: Bearer ENTERPRISE_API_KEY' \
--header 'Content-Type: application/json'
```

## Body

```json json theme={"theme":{"light":"github-light","dark":"github-dark"}}
{
  "model": "default",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "What is the capital of France?"}
  ],
  "max_tokens": 500,
  "temperature": 0.7,
  "stream": false
}
```

## Body Attributes

The body follows the OpenAI Chat Completions format. The server sends every field to the model without changes, except `model`, `models`, `route`, and `preset`.

<ParamField body="model" type="string" required>
  Required by most SDKs. Send any value: the server always uses your plan's model.
</ParamField>

<ParamField body="messages" type="array" required>
  The conversation, as a list of `{ "role", "content" }` objects. Roles are `system`, `user`, `assistant`, and `tool`. `content` can be a string or a list of content parts, such as text parts with `cache_control` or image parts if the model accepts images.
</ParamField>

<ParamField body="stream" type="boolean" default="false">
  Set to `true` to get the reply as Server-Sent Events while the model generates it. Use streaming for long outputs: a request that is not streamed must finish in about 110 seconds.
</ParamField>

<ParamField body="max_tokens" type="integer">
  The maximum number of tokens in the reply.
</ParamField>

<ParamField body="temperature" type="number">
  The sampling temperature. Higher values make the output more random.
</ParamField>

<ParamField body="top_p" type="number">
  Nucleus sampling: the model picks the next token only from tokens whose total probability is `top_p`.
</ParamField>

<ParamField body="tools" type="array">
  Functions the model can call, in the OpenAI `tools` format. Use `tool_choice` to control when the model calls them.
</ParamField>

<ParamField body="response_format" type="object">
  Ask for JSON output, for example `{ "type": "json_object" }` or a `json_schema`, if the model supports it.
</ParamField>

<ParamField body="stop" type="string | array">
  One or more sequences where the model stops.
</ParamField>

## Response

```json json theme={"theme":{"light":"github-light","dark":"github-dark"}}
{
  "id": "gen-abc123",
  "object": "chat.completion",
  "created": 1790000000,
  "model": "xiaomi/mimo-v2.6-pro",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "The capital of France is Paris."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 25,
    "completion_tokens": 8,
    "total_tokens": 33,
    "prompt_tokens_details": { "cached_tokens": 0 }
  }
}
```

## Streaming

With `"stream": true`, the response is `text/event-stream`. Each event is a `data:` line that holds one chunk, and the stream ends with `data: [DONE]`.

<CodeGroup>
  ```python Python theme={"theme":{"light":"github-light","dark":"github-dark"}}
  from openai import OpenAI

  client = OpenAI(
      base_url="https://modelslab.com/api/v1/enterprise/proxy/v1",
      api_key="YOUR_ENTERPRISE_API_KEY",
  )

  stream = client.chat.completions.create(
      model="default",
      messages=[{"role": "user", "content": "Tell me a short story."}],
      stream=True,
  )
  for chunk in stream:
      if chunk.choices and chunk.choices[0].delta.content:
          print(chunk.choices[0].delta.content, end="", flush=True)
  ```

  ```bash cURL theme={"theme":{"light":"github-light","dark":"github-dark"}}
  curl -N https://modelslab.com/api/v1/enterprise/proxy/v1/chat/completions \
    -H "Authorization: Bearer $ENTERPRISE_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{"model": "default", "stream": true, "messages": [{"role": "user", "content": "Tell me a short story."}]}'
  ```
</CodeGroup>

```text theme={"theme":{"light":"github-light","dark":"github-dark"}}
data: {"id":"gen-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":"Once"}}]}

data: {"id":"gen-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":" upon"}}]}

data: [DONE]
```

## Prompt caching

If the model supports prompt caching, mark a long, repeated part of the prompt with `cache_control`. The next requests that start with the same prefix read it from the cache. `usage.prompt_tokens_details.cached_tokens` shows how many tokens came from the cache.

```json json theme={"theme":{"light":"github-light","dark":"github-dark"}}
{
  "model": "default",
  "messages": [
    {
      "role": "system",
      "content": [
        {"type": "text", "text": "<long reference document>", "cache_control": {"type": "ephemeral"}}
      ]
    },
    {"role": "user", "content": "Summarise section 2."}
  ]
}
```