<!-- Generated from openapi/hiapi.json by scripts/generate-llms.mjs. Do not edit manually. -->
# HiAPI

> HiAPI provides Responses and Chat Completions for text models and an asynchronous task API for image, video, and audio generation.

API base URL: https://api.hiapi.ai
Documentation: https://www.hiapi.ai/docs/

## Choose an integration

- Text API: use Responses or compatible Chat Completions according to the model page.
- Media API: use the asynchronous task workflow for image, video, and audio models.
- Remote MCP: connect to https://mcp.hiapi.ai/mcp with the API key as a Bearer token. It exposes 9 live-catalog tools: `generate_image`, `edit_image`, `generate_video`, `generate_audio`, `generate_text`, `get_task_status`, `list_models`, `get_model_capabilities`, and `get_pricing`. Media generation uses `/v1/tasks`; text generation uses the model's Chat Completions or Responses route.
- Agent Skills: browse https://github.com/HiAPIAI/hiapi-skills.

## Authentication

```http
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json
```

Create or rotate keys at https://www.hiapi.ai/en/dashboard/api-keys.

## Create a response

`POST https://api.hiapi.ai/v1/responses`

Use this endpoint only for models whose page lists Responses support. Required fields are `model` and `input`; reasoning, storage, tools, and streaming semantics remain model-specific.

```json
{
  "model": "deepseek-v4-pro",
  "input": "Explain the most important tradeoff.",
  "reasoning": {
    "effort": "high"
  },
  "stream": false
}
```

## Create a chat completion

`POST https://api.hiapi.ai/v1/chat/completions`

Required fields are `model` and `messages`. Set `stream: true` for Server-Sent Events. Read the answer from `choices[0].message.content` and token billing details from `usage`.

```json
{
  "model": "deepseek-v4-flash",
  "messages": [
    {
      "role": "user",
      "content": "Explain sparse attention in three sentences."
    }
  ],
  "stream": false
}
```

## Create a generation task

`POST https://api.hiapi.ai/v1/tasks`

- `model` (string, required): HiAPI model ID.
- `route` (string, optional): model route, such as beta, pro, or ext.
- `input` (object, required): Model-specific parameters. See the selected model page for its input schema.
- `callback` (object, optional): Terminal delivery settings. HTTP or HTTPS endpoint that receives the terminal task payload.
- `storage` (string, optional): Requested output storage tier.

Use an `Idempotency-Key` header when a submission may be retried. Reusing the same key with the same request returns the original `taskId`.

```json
{
  "model": "gpt-image-2/text-to-image",
  "input": {
    "prompt": "A premium product photo on a clean studio background",
    "aspect_ratio": "1:1",
    "resolution": "1K"
  },
  "callback": {
    "url": "https://example.com/hiapi/callback",
    "when": "final"
  }
}
```

The response returns `data.taskId` immediately.

## Get the result

Poll `GET https://api.hiapi.ai/v1/tasks/{taskId}`, or provide `callback.url` when creating the task.

Interim status: `queued` → `handling` → `archiving`. Terminal status: `success` or `fail`.

On `success`, read generated files from `data.output[].url`. Use exponential backoff when polling.

## Find a model and its input fields

First fetch the machine-readable model index:

`https://www.hiapi.ai/docs/models.json`

Use each entry's `summary`, `capability`, and `bestFor` to select a model. Only call entries with `availability: "online"`. Text models expose `request`; media models expose `input` and optional `routes`.

Then open the selected entry's `docs` or `docsZh` URL for full field descriptions, examples, limits, and pricing links. Do not guess model names or parameters.

- API documentation model directory: https://www.hiapi.ai/docs/models/
- Image API collection: https://www.hiapi.ai/docs/collections/ai-image-apis/
- Video API collection: https://www.hiapi.ai/docs/collections/ai-video-apis/
- Audio API collection: https://www.hiapi.ai/docs/collections/ai-audio-apis/
- Text API collection: https://www.hiapi.ai/docs/collections/ai-text-apis/

## Sources of truth

- Unified Async API: https://www.hiapi.ai/docs/async-api/
- Responses API: https://www.hiapi.ai/docs/text-api/responses/
- Chat Completions API: https://www.hiapi.ai/docs/text-api/chat-completions/
- OpenAPI: https://www.hiapi.ai/docs/hiapi.json
- Pricing: https://www.hiapi.ai/en/pricing
- Changelog: https://www.hiapi.ai/docs/changelog/