AIAgentConfig

The ASR language ID. Valid values:

• zh_mandarin: Chinese.

• en: English.

• zh_en: Mixed Chinese and English.

• es: Spanish.

• jp: Japanese.

The silence duration threshold for speech segmentation. Silence exceeding this threshold is considered a sentence break. Valid values: 200 to 1200. Unit: milliseconds. Default value: 400.

The interruption threshold parameter. Valid values: 0 to 11. Default value: 11.

• 0: Disables the voice activity detection (VAD) feature.

• 1 to 10: A higher value indicates that it is more difficult to interrupt the agent.

• 11: Significantly different from previous values, providing lower speech distortion during pre-processing and stronger anti-interference capability.

The pass-through parameters for the proprietary ASR integration.

The minimum duration threshold for voice activity detection, which is used to control the interruption sensitivity. A value of 0 disables this feature. Valid values: 200 to 2000. Unit: milliseconds. Common values: 200 to 500, which correspond to 1 to 4 words. Default value: empty (does not take effect).

The voice ID. Changes take effect in the next sentence. If not specified, the voice ID configured in the agent template is used. This parameter takes effect only for preset TTS. Maximum length: 64 characters. For valid values, refer to Intelligent speech effect examples.

The model ID. Currently, only minimax is supported. Valid values: speech-01-turbo or speech-02-turbo.

The language ID. Currently, only minimax is supported. Default value: empty. This parameter enhances the recognition capability for specified minority languages and dialects. After configuration, it improves speech performance in specified minority language or dialect scenarios. If the minority language type is unclear, you can set this parameter to "auto", and the model will automatically determine the language type. Valid values:

The emotion. Currently, only minimax is supported. minimax supports the following 7 emotions:

• happy: Happy

• sad: Sad

• angry: Angry

• fearful: Fearful

• disgusted: Disgusted

• surprised: Surprised

• calm: Calm

The speech rate. Supported on all platforms. For cosyvoice, the default value is 1.0, and the valid range is 0.5 to 2.0. For minimax, the default value is 1.0, and the valid range is 0.5 to 2.0.

The maximum number of retained turns for the historical conversation context of the LLM or MLLM. Default value: 10.

The system prompt for the LLM after the call starts.

The parameters for Alibaba Cloud Model Studio Application Center. The type is a JSON string. For the parameter format, refer to: Alibaba Cloud Model Studio Application Center parameters.

The additional query parameters for the OpenAI-compatible LLM. Parameters must be in the key=value format. Multiple parameters are connected by an ampersand (&). All values must be of the string type.

After this feature is enabled, the agent sends the complete LLM result to the client after the LLM generates the full response. This switch does not affect the streaming generation of subtitles.

The minimum length of text output in characters. Text shorter than this length is cached and waiting for concatenation. Valid values: 0 to 100. A value of 0 or empty indicates no limit. Default value: empty.

The maximum delay time for text output in milliseconds. If this time is exceeded, the cached text is forcibly output. Valid values: 1000 to 10000. A value of 0 or empty indicates no limit. Default value: empty.

Specifies whether the LLM message history is consistent with the TTS playback content. Default value: false. After this feature is enabled, the saved LLM messages are consistent with the TTS playback content.

[
  {"role": "user", "content": "Tell me a story."},
  {"role": "assistant", "content": "Sure, let me tell you a story about the Romance of the Three Kingdoms. Do you<ims_agent_interrupted> want to listen?"},
  {"role": "user", "content": "Let's change to another one."}
]

The model ID of the digital human.

Specifies whether to support voice interruption. Default value: true.

The ASR processing policy in mode.

• cache: Caches the ASR text. After the current turn ends, the cached ASR text is processed together in the next turn.

• discard: Discards the ASR text directly.

The default policy is to cache the ASR text.

Specifies whether to retain interruption words when sending text to the LLM. Default value: discard.

Specifies whether to enable voiceprint recognition. Default value: false. When voiceprint recognition is enabled, a valid voiceprint ID must be passed.

The unique identity ID for voiceprint recognition. Default value: empty. The passed voiceprint ID must have been registered through the voiceprint registration API. For the API documentation, refer to Register voiceprint.

The voiceprint registration mode. Default value: Explicit.

The turn detection mode.

• Normal (default): Normal mode. AI is not used to determine semantics.

• Semantic: Semantic mode. AI is used to determine whether the user has finished speaking based on the context semantics.

The pause determination time in AI mode. Unit: milliseconds. Default value: -1.

• -1: AI automatically determines the appropriate waiting time.

• 0 to 10000: Custom waiting time. We recommend that you set it to 0 to 1500 milliseconds.

The eagerness level. This parameter takes effect only in Semantic mode. It controls the response speed after the AI detects a pause:

• "Low": Patient waiting. The maximum waiting time is 6 seconds, which reduces the risk of being interrupted.

• "Medium": Balanced waiting. The maximum waiting time is 4 seconds, which is suitable for most scenarios.

• "High": Fast response. The maximum waiting time is 2 seconds, which improves speed but may increase the risk of false cuts.

Default value: empty.

The ID of the ambient sound for calls. You can obtain this ID in the advanced settings of the agent configuration in the console.

The volume of the background sound for calls. Valid values: 0 to 100. A value of 0 indicates that the background sound is disabled.