⌘K

Tongyi Intent Detect V3

tongyi-intent-detect-v3

Intent understanding model that quickly and accurately parses user intent within milliseconds and selects appropriate tools to solve problems. Supports three modes: intent + function call, intent only, and function call only.

Authentication

authorization string required

All APIs require authentication via Bearer Token.

Get API Key:

Visit API Key Management Page to get your API Key.

Usage:

Add to request header:

Authorization: Bearer YOUR_API_KEY

Parameters

model string required

Model ID to use for the request.

Value: tongyi-intent-detect-v3

messages array required

Array of message objects representing the conversation history.

role string required

Message role.

Options: user, assistant, system

content string required

Message content. System message must include mode instructions and configuration.

max_tokens integer

Maximum tokens to generate in the completion.

Range: 1 - 1024

temperature number

Sampling temperature to use.

Default: 1.0

Range: 0.0 - 2.0

Usage Modes

Intent + Function Call Mode (INTENT_MODE)

Outputs both intent tags and function call information. System message format:

You are Qwen, created by Alibaba Cloud. You are a helpful assistant. You may call one or more tools to assist with the user query. The tools you can use are as follows:
[Tool definitions in JSON format]
Response in INTENT_MODE.

Response Format:

<tags>
[intent tags]
</tags><tool_call>
[function call array]
</tool_call><content>
response text
</content>

Intent Only Mode

Outputs only intent classification. System message format:

You are Qwen, created by Alibaba Cloud. You are a helpful assistant.
You should choose one tag from the tag list:
{intent dictionary in JSON format}
Just reply with the chosen tag.

Tips: Use single uppercase letters (A, B, C…) as intent keys for faster response (always 1 token).

Function Call Only Mode (NORMAL_MODE)

Outputs only function call information. System message format:

You are Qwen, created by Alibaba Cloud. You are a helpful assistant. You may call one or more tools to assist with the user query. The tools you can use are as follows:
[Tool definitions in JSON format]
Response in NORMAL_MODE.

Response Format:

<tool_call>
{function call object}
</tool_call>

Tool Definition Format

Tools should be defined in the system message as a JSON array:

[
  {
    "name": "function_name",
    "description": "Function description",
    "parameters": {
      "type": "object",
      "properties": {
        "param_name": {
          "type": "string",
          "description": "Parameter description"
        }
      },
      "required": ["param_name"]
    }
  }
]

Multi-turn Conversations

The model supports multi-turn conversations. If insufficient information is provided, it will ask follow-up questions before making function calls.

Example:

User: “I want to check the weather”
Model: “Which city would you like to check?”
User: “Hangzhou”
Model: Returns function call with location parameter

Response Format

Standard chat completion response with model-specific content formatting based on the selected mode.

id string

Unique identifier for the completion.

choices array

Array of completion choices.

message object

The generated message.

content string

Response content in the format specified by the mode (intent tag, XML-formatted tool calls, or plain text).

usage object

Token usage statistics.

prompt_tokens integer

Number of tokens in the prompt.

completion_tokens integer

Number of tokens in the completion.

total_tokens integer

Total number of tokens used.