Dify

Dify is an open-source LLM application platform that lets you build AI workflows with a visual drag-and-drop editor — no code required for prompt orchestration, RAG pipelines, or tool integrations. Choose Dify if you need complex AI agent logic (knowledge bases, multi-step workflows, tool calling) managed outside your code.
Usage
To use Dify as the LLM engine, pass the following JSON in the LLMConfig field of the StartAIConversation API. Dify uses its own dedicated LLMType:
// json — LLMConfig
{
  "LLMType": "dify",
  "APIKey": "app-<your_dify_api_key>",
  "APIUrl": "https://api.dify.ai/v1/chat-messages",
  "Streaming": true,
  "Inputs": {},
  "User": "user_id",
  "ConversationId": ""
}
For full Conversational AI configuration (STT, TTS, interruption handling, VAD, etc.), see the TRTC Conversational AI API Reference.
Parameter reference
Field
Type
Required
Description
LLMType
String
Yes
Fixed value: "dify".
APIKey
String
Yes
Dify app API key (starts with app-). Obtain from your Dify app settings.
APIUrl
String
Yes
Dify chat messages endpoint.
Streaming
Boolean
No
Enable streaming. Default: true.
Inputs
Object
No
Input variables defined in your Dify app workflow.
User
String
No
User identifier for Dify session tracking.
ConversationId
String
No
Leave empty for a new conversation; pass an existing ID to continue.
For more details on Dify platform, see the Dify documentation.
Next step: Configure your TTS provider﻿

Field	Type	Required	Description
`LLMType`	String	Yes	Fixed value: `"dify"`.
`APIKey`	String	Yes	Dify app API key (starts with `app-`). Obtain from your Dify app settings.
`APIUrl`	String	Yes	Dify chat messages endpoint.
`Streaming`	Boolean	No	Enable streaming. Default: `true`.
`Inputs`	Object	No	Input variables defined in your Dify app workflow.
`User`	String	No	User identifier for Dify session tracking.
`ConversationId`	String	No	Leave empty for a new conversation; pass an existing ID to continue.