Skip to content

README.md

Latest commit

 

History

History
73 lines (53 loc) · 4.1 KB

README.md

File metadata and controls

73 lines (53 loc) · 4.1 KB
Raw

Channel Integration Guide

Connect CodeMux to your favorite messaging platforms to interact with AI coding engines directly from IM apps — with real-time streaming, tool execution, and multi-engine access.

Supported Platforms

Platform Connection Streaming Group Creation Rich Content Guide
Feishu / Lark WebSocket SDK ✅ Edit-in-place ✅ Auto-create Interactive Cards → Setup
DingTalk (钉钉) Stream mode (WS) ✅ AI Card ✅ Scene groups ActionCard / Markdown → Setup
Telegram Webhook / Long Polling ✅ Draft / Edit ❌ P2P only MarkdownV2 + Buttons → Setup
WeCom (企业微信) HTTP Callback (AES XML) ❌ Batch mode ✅ App group chat Markdown → Setup
Microsoft Teams Bot Framework HTTP ✅ Edit-in-place ❌ P2P only Adaptive Cards v1.5 → Setup
WeChat iLink (微信个人号) HTTP long-poll ❌ Batch mode ❌ P2P only Plain text → Setup

Connection Types

Type Platforms Tunnel Required How It Works
Direct Connect Feishu / Lark, DingTalk, Telegram (polling) No Platform SDK maintains persistent connection from CodeMux to the platform's servers
Webhook WeCom, Teams, Telegram (webhook) Yes Platform sends HTTP requests to your CodeMux instance via Cloudflare Tunnel

Session Models

Model Platforms Flow
One Group = One Session Feishu / Lark, DingTalk, WeCom P2P chat → select project → group auto-created → all messages in that group go to one CodeMux session
P2P Direct Telegram, Teams, WeChat iLink Interact directly in private chat with temporary sessions. In group chats (Telegram / Teams), @mention the bot

Common Features

All channels support the same unified slash-command surface (powered by electron/main/channels/shared). Streaming and rich-content support varies by platform — see the table above.

Common Commands Reference

Commands are split into two groups by capability. Channels expose the right subset based on the chat context (P2P vs group).

P2P chat (entry point) — full command set

Command Description
/project Show project list (reply with a number to switch)
/new Create a new session in the current project (auto-creates a group on Feishu/DingTalk/WeCom)
/switch Show session list for the current project (reply with a number to open)
/cancel Cancel the in-flight AI request
/status Show current session info
/mode <agent|plan|build> Switch agent mode
/model [list|<id>] List available models, or switch to a specific id
/history Show recent message history
/help Show all available commands
/start Same as /help (welcome alias used by Telegram / Teams)

Group / channel chat — session-ops only

In a session group (Feishu/DingTalk/WeCom auto-created group, or a Telegram/Teams group with the bot @mentioned), navigation commands are hidden because the group is the session. Available commands:

/cancel, /status, /mode, /model, /history, /help

In Telegram and Teams groups you must @mention the bot or prefix with /command for the bot to receive the message. In Feishu / DingTalk / WeCom session groups, every message is forwarded to the AI engine automatically.

Architecture

IM Platform (Feishu/Lark/DingTalk/Telegram/WeCom/Teams/WeChat iLink)
  ↕ Messages (SDK WebSocket or HTTP Webhook)
Channel Adapter (in CodeMux)
  ↕ WebSocket (internal)
Gateway Server
  ↕ Engine routing
AI Engine (OpenCode / Copilot / Claude Code / Codex experimental)

Each channel adapter connects to the CodeMux gateway using the same WebSocket protocol as the web UI, ensuring feature parity.