AI Model APIs
This page summarizes the AI model endpoints in the API section. Regular clients, plugins, and scripts usually start with OpenAI-compatible endpoints such as /v1/chat/completions, /v1/responses, and /v1/models.
Recommended entry
If this is your first integration, read [Authentication](/en/api/authentication) first, then start testing with [ChatCompletions Format](/en/api/ai-model/chat/openai/createchatcompletion).
Endpoint categories
| Category | Description | Recommended entry |
|---|---|---|
| Models | Fetch models available to the current account | List Models |
| Chat | Conversational generation, the most common model call | ChatCompletions Format |
| Responses | OpenAI Responses-style endpoint | Responses Format |
| Claude Messages | Claude Messages API format | Claude Messages Format |
| Completions | Traditional text completion endpoint | Create Completion |
| Embeddings | Text vectorization for retrieval and RAG | Create Embedding |
| Rerank | Rerank documents or candidate results | Create Rerank |
| Moderations | Content safety moderation | Create Moderation |
| Audio | Speech synthesis, transcription, and translation | Create Speech |
| Realtime | Realtime audio session endpoint | Create Realtime Session |
| Images | Image generation and editing | Image Generation |
| Videos | Video generation and task status | Create Video |
Before you call
- Make sure the API address includes
/v1, such ashttps://api.dli.li/v1. - Make sure the request header includes
Authorization: Bearer sk-your-api-key. - Make sure the model name is available to the current account.
- If a third-party client only supports the OpenAI protocol, prefer the OpenAI-compatible endpoints.