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

# Chat and Responses

> Generate text, tool calls, and Responses-style outputs through native paid inference.

Use these routes when you want native Compose receipts, raw x402 support, or Compose Key billing.

## Routes

| Method | Path                             | Streaming | Description                                                                 |
| ------ | -------------------------------- | --------- | --------------------------------------------------------------------------- |
| `POST` | `/v1/chat/completions`           | Yes       | OpenAI Chat Completions-shaped request and response.                        |
| `POST` | `/v1/responses`                  | Yes       | Responses-shaped input, tools, multimodal content, and stored response IDs. |
| `GET`  | `/v1/responses/{id}`             | No        | Retrieve a stored response record.                                          |
| `GET`  | `/v1/responses/{id}/input_items` | No        | Return stored input items for a response.                                   |
| `POST` | `/v1/responses/{id}/cancel`      | No        | Cancel a stored response or async media job.                                |
| `POST` | `/api/inference`                 | Yes       | Legacy alias for `/v1/chat/completions`.                                    |

## Headers

<ParamField header="Authorization" type="string">
  `Bearer compose-...` for Compose Key billing.
</ParamField>

<ParamField header="PAYMENT-SIGNATURE" type="string">
  Raw x402 payment signature for x402-aware clients.
</ParamField>

<ParamField header="x-x402-max-amount-wei" type="string">
  Maximum USDC atomic amount for raw x402 `upto` payments.
</ParamField>

## Chat body

<ParamField body="model" type="string" required>
  Public catalog model ID.
</ParamField>

<ParamField body="messages" type="array" required>
  Non-empty OpenAI-style chat message array.
</ParamField>

<ParamField body="stream" type="boolean">
  Enables SSE streaming. Non-boolean values return `400 invalid_stream`.
</ParamField>

<ParamField body="tools" type="array">
  OpenAI function tool definitions for models that support tool calling.
</ParamField>

<ParamField body="tool_choice" type="string | object">
  `none`, `auto`, `required`, or a named function choice.
</ParamField>

<ParamField body="response_format" type="object">
  Text, JSON object, or JSON schema response format when the provider supports it.
</ParamField>

## Responses body

<ParamField body="model" type="string" required>
  Public catalog model ID.
</ParamField>

<ParamField body="input" type="string | array" required>
  Responses-style input. Required unless `previous_response_id` supplies history.
</ParamField>

<ParamField body="previous_response_id" type="string">
  Response ID to hydrate prior messages.
</ParamField>

<ParamField body="modalities" type="array">
  Requested output modalities.
</ParamField>

## Streaming events

Native streams send OpenAI-compatible data frames plus Compose events:

| Event                    | Description                                     |
| ------------------------ | ----------------------------------------------- |
| `data: {...}`            | Chat or Responses chunk.                        |
| `event: compose.receipt` | In-band settlement receipt before stream close. |
| `event: compose.error`   | Structured Compose error frame.                 |
| `data: [DONE]`           | Stream terminator.                              |