# ModelVerse OpenAPI

This document lists the APIs supported by ModelVerse and explains the custom extension fields.

## Supported Protocol Endpoints

| Protocol/API | Description | Example Path |
| --- | --- | --- |
| OpenAI Chat Completions | Text dialogue/streaming output | `POST /v1/chat/completions` |
| OpenAI Responses | Multi-part output and tool call | `POST /v1/responses` |
| OpenAI Images | Image generation, editing | `POST /v1/images/generations`, `/v1/images/edits` |
| OpenAI Audio | Text-to-speech | `POST /v1/audio/speech` |
| Model Enumeration (OpenAI/Gemini) | Returns a list of available models | `GET /v1/models`, `GET /v1beta/models` |
| Anthropic Messages | Claude protocol dialogue | `POST /v1/messages` |
| Google Gemini | Content generation in `/v1beta/models/{model}:{action}` style | `POST /v1beta/models/*` |
| Asynchronous Tasks | Video/long-duration task submission and query | `POST /v1/tasks/submit`, `GET /v1/tasks/status` |

## Custom Extension Fields

On top of compatibility with the official OpenAI/Gemini protocols, we provide a few optional fields that are only recognized when using ModelVerse OpenAPI; you do not need to fill these when directly calling official OpenAI services.

| Field | Applicable Interface | Description | Usage |
| --- | --- | --- | --- |
| `web_search` | `POST /v1/chat/completions` | Controls whether to enable internet search capabilities and choose a search provider. Disabled by default. Supported `vendor`: `tencent_sougo` (Tencent Sogou, default), `bocha` (Bocha API), `bing`. | Add at the root node of the request JSON:<br/>```json<br/>"web_search": {<br/>  "enable": true,<br/>  "vendor": "tencent_sougo"<br/>}``` |
| `thinking_enabled` | `POST /v1/chat/completions` | Internal "thinking mode" switch, affects model-side inference strategy. Disabled by default. | Submit alongside the OpenAI request:<br/>`"thinking_enabled": true` |