Get assistant details

Path parameter · Assistant ID

Unique assistant ID

Owning project ID

Assistant name

Currently active version ID

Assistant scene type, e.g. inbound_knowledge (inbound knowledge base) or outbound_collect (outbound collection)

Assistant objective description

Text-to-speech (TTS) voice ID

Text-to-speech (TTS) voice name

WebSocket connection URL

WebSocket authentication token

Speech recognition (ASR) model ID

Speech recognition (ASR) model name

Greeting/opening line; the first text the assistant plays once the call is connected

System prompt defining the assistant's role, behavior and conversation strategy

Data collection config; defines the list of questions the assistant collects during the conversation

Question to collect

Preset answer options (optional)

Associated knowledge base code (RAG knowledge base identifier)

Hot word list used to improve ASR accuracy

Hot word text

Hot word weight; higher values take priority

Homophones/near-homophones this hot word can replace

Whether fuzzy matching is enabled for this hot word

Total number of calls made by the assistant

Creator user ID

创建者用户信息

User ID

User email

Username

Whether the record uses the legacy format (compatibility flag)

Associated custom LLM configuration ID

Whether fuzzy matching is enabled (for ASR error correction)

Feature flags (comma-separated identifiers)

Time zone used by the assistant, e.g. Asia/Shanghai

Conversation flow definition (flow chart config in JSON)

打断配置，控制用户在助手说话时如何打断

Interruption method

Uninterruptible duration after playback starts (ms); user speech within this window does not trigger an interruption

Talk-over detection threshold (ms); an interruption triggers only after the user speaks longer than this

Whether to resume the interrupted playback after an interruption

端点检测配置（End of Utterance），用于判断用户是否说完

Whether smart end-of-utterance detection is enabled

Wait timeout (ms); the utterance is considered finished if no speech is detected within this time

Unfinished utterance timeout (ms); extra wait time when an unfinished utterance is detected

VAD（Voice Activity Detection）语音活动检测配置

Silence padding duration (ms); extra wait time after speech ends

Speech detection sensitivity ratio

VAD mode (0-3); higher values are stricter

Positive speech threshold; values above it are judged as speech

Negative speech threshold; values below it are judged as no speech

Redemption frames; extra frames checked before declaring speech ended

Minimum speech frames; segments shorter than this are ignored

Energy threshold; audio below this energy is judged as silence

Speech start detection duration (ms)

Speech end detection duration (ms)

AI Agent 核心配置，包括对话策略和LLM模型参数

Slice time (ms); interval used to split long speech for processing

Silent wait time after an interruption (ms)

Maximum number of follow-up prompts when the user stays silent

Maximum number of dialogue gaps; the call ends automatically once exceeded

Maximum session duration (seconds); the call ends automatically once exceeded

Whether the opening line can be interrupted

Speech interval threshold (ms); two utterances closer than this are treated as continuous speech

Whether the assistant stays silent while VAD detects the user speaking

LLM model name, e.g. gpt-4o, claude-sonnet-4.5

LLM API base URL

LLM API authentication token

LLM vendor identifier, e.g. openai, azure, volcengine

API version (mainly for Azure OpenAI)

LLM temperature (0-2); higher values produce more random output

LLM Top-P sampling parameter (0-1)

Prompt mode, e.g. system (system prompt)

Agent decision threshold

Agent run mode

Whether to play filler words (e.g. 'hmm', 'okay') while the AI is thinking

Filler word list

Function calling definitions (OpenAI format)

Function call hooks; run extra logic before/after a tool call

Target tool name

Hook prompt; extra instructions injected when the tool is called

Knowledge base code (for RAG retrieval)

Number of top-K entries returned by knowledge base retrieval

SCRM（社交化客户关系管理）集成配置

SCRM bot ID

SCRM app key

SCRM app secret

技能配置

按键技能配置（DTMF按键检测）

Maximum idle wait time (seconds); key-press listening stops once exceeded

短时记忆配置，用于对话中的上下文摘要

Points of interest to remember

LLM模型配置（通用）

Model vendor, e.g. openai, azure, volcengine

Model name

API base URL

API authentication token

Temperature parameter

API version

Top-P sampling parameter

知识库检索配置

Use the previous N dialogue rounds as retrieval context

Minimum match score for knowledge base chunks; results below this score are filtered out

Whether to use memory-enhanced retrieval queries

Whether knowledge base retrieval is disabled at the start of a new dialogue round

Preset speech scripts; key is the scene identifier, value is the script content

音频处理配置（降噪、增益等）

VAD engine type

Whether noise suppression is enabled

Noise suppression algorithm: rnnoise or dfn

Noise suppression intensity level

Whether automatic gain control (AGC) is enabled

Whether all audio processing is disabled (noise suppression + AGC)

语音合成（TTS）播报配置

Speech rate multiplier; 1.0 is normal, above 1 is faster, below 1 is slower

TTS engine name

Extra TTS engine options

Whether SSML (Speech Synthesis Markup Language) is enabled

SSML text replacement rules

Original text to replace

Replacement SSML markup text

全局配置

Global greeting

Global prompt

Scene type

Primary language, e.g. zh-CN, en-US, ja-JP

多语言配置

Additional supported languages

Language code, e.g. en-US, ja-JP

Debug mode flag

Call direction: inbound or outbound

Whether the dialog cannot be interrupted

Prompt mode

工作时间配置

Work time rule ID

通话分析配置

Whether real-time data collection is enabled (extracts key information during the call)

Collection target definitions

Target name

Enum value list (restricts allowed values)

Owning branch identifier

Target summary

Unique identifier

Conversation analysis dimensions

Analysis target identifier

Analysis dimension description

通话质量分析配置

Quality analysis prompt/rule content

MCP server configuration list, used to extend the assistant's tool capabilities

MCP server name

Server description

Server type, e.g. stdio or sse

Start command

Start arguments

Environment variables

Timeout (seconds)

Selected tools

Tool name

Whether the tool runs asynchronously

音轨配置（背景音、音量控制）

Master volume (default 1.0)

Main track volume (default 0.8), i.e. the volume of the TTS voice

Effect audio tracks (background music, etc.)

Audio file name/path

Volume (default 0.8)

Whether the track is enabled

Playback mode, e.g. loop or once

查询重写配置，用于优化知识库检索的查询语句

Whether query rewriting is enabled

Rewrite prompt

LLM模型配置（通用）

Model vendor, e.g. openai, azure, volcengine

Model name

API base URL

API authentication token

Temperature parameter

API version

Top-P sampling parameter

语音识别（ASR）高级配置

ASR provider list (multiple providers can recognize in parallel)

ASR model ID

Provider name

Extra provider configuration options

ASR result submission mode

Whether transcription ASR is enabled (transcribes the entire call)

Fuzzy matching options

Recognition direction

Creation time

Last update time

Soft-delete time; null if not deleted