Feedback

Large Language Model Configuration

Once user speech is transcribed into text, the LLM module processes the input and generates a contextually relevant response. TRTC's flexible framework lets you plug in any leading LLM — including OpenAI-compatible models and agent platforms such as Dify and Coze — so you can pick the engine that best fits your use case. With end-to-end conversation latency under 1 second globally, replies feel instant to end users. Simply configure the connection through the LLMConfig object, specifying your model, API credentials, and optional parameters such as system prompt and conversation history depth.
Available Providers
Provider
Models
Integration
Best for
﻿OpenAI﻿
gpt-5.5, gpt-5.4, gpt-4o-mini etc.
OpenAI-compatible
Top-tier reasoning, multimodal
﻿Gemini﻿
gemini-3.1-pro-preview, gemini-3-flash-previewetc.
OpenAI-compatible
Large context, cost-efficient
﻿MiniMax﻿
MiniMax-M2.7, MiniMax-M2.5etc.
OpenAI-compatible
Chinese language, competitive pricing
﻿Tencent Hunyuan﻿
hunyuan-2.0-thinking-20251109, hunyuan-2.0-instruct-20251111etc.
OpenAI-compatible
Tencent Cloud ecosystem, Cost-competitive
﻿Dify﻿
Any LLM (configurable)
Platform
No-code AI workflows, RAG
﻿Coze﻿
Any LLM (configurable)
Platform
No-code bot builder, plugins
Note:
OpenAI-compatible providers (OpenAI, Gemini, MiniMax, Hunyuan) share a unified configuration structure. Platform providers (Dify, Coze) have their own dedicated config — see each sub-page for details.
For the complete LLM parameter reference, see the TRTC Conversational AI API Reference.
LLM HTTP Request Headers
TRTC automatically adds the following headers to each LLM request, which can be used for logging, routing, or debugging:
X-Task-Id: <task_id_value>       // Task ID
X-Request-Id: <request_id>       // Request ID (same for retries)
X-Sdk-App-Id: SdkAppId
X-User-Id: UserId
X-Room-Id: RoomId
X-Room-Id-Type: "0"              // "0" = numeric, "1" = string
Header Reference
Header
Description
X-Task-Id
Unique task identifier for the current AI conversation session.
X-Request-Id
Request identifier. Remains the same across retries for the same request.
X-Sdk-App-Id
Your TRTC application's SdkAppId.
X-User-Id
The user ID in the current conversation.
X-Room-Id
The room ID of the current TRTC session.
X-Room-Id-Type
Room ID type. "0" = numeric, "1" = string.
Note:
These headers are useful when you want to track or route LLM requests through a proxy or logging service.

Provider	Models	Integration	Best for
OpenAI	`gpt-5.5`, `gpt-5.4`, `gpt-4o-mini` etc.	OpenAI-compatible	Top-tier reasoning, multimodal
Gemini	`gemini-3.1-pro-preview`, `gemini-3-flash-preview`etc.	OpenAI-compatible	Large context, cost-efficient
MiniMax	`MiniMax-M2.7`, `MiniMax-M2.5`etc.	OpenAI-compatible	Chinese language, competitive pricing
Tencent Hunyuan	`hunyuan-2.0-thinking-20251109`, `hunyuan-2.0-instruct-20251111`etc.	OpenAI-compatible	Tencent Cloud ecosystem, Cost-competitive
Dify	Any LLM (configurable)	Platform	No-code AI workflows, RAG
Coze	Any LLM (configurable)	Platform	No-code bot builder, plugins

Header	Description
`X-Task-Id`	Unique task identifier for the current AI conversation session.
`X-Request-Id`	Request identifier. Remains the same across retries for the same request.
`X-Sdk-App-Id`	Your TRTC application's SdkAppId.
`X-User-Id`	The user ID in the current conversation.
`X-Room-Id`	The room ID of the current TRTC session.
`X-Room-Id-Type`	Room ID type. `"0"` = numeric, `"1"` = string.