Chatwoot AI Agent — Multi-Inbox Intelligent Customer Service

v1.0 — The Ancestor Version
The foundation of a multi-tenant AI customer service platform built on self-hosted Chatwoot.

Architecture Overview

Customer (via Web Widget)
        │
        ▼
Chatwoot Self-Hosted (wss://)
        │
        ▼
Chatwoot WS Agent ←── INBOX_CONFIG routing
   │          │
   ▼          ▼
sourcing-agent  halo-blog-agent
 (Inbox 1)      (Inbox 7)
 English        简体中文

Core Features

Feature Description
7×24 AI Auto-Reply AI responds instantly to customer inquiries
AI ↔ Human Handoff Seamless switching: human replies → AI backs off; human leaves → AI resumes (15min timeout)
Customer-Unaware AI uses User session, not AgentBot — customer sees "agent reply"
Multi-Inbox Routing Single WS agent routes by inbox_id to different AI agents & knowledge bases
Private Notes AI auto-generates Chinese translation notes for human agents
Offline Resilience Bubble stays visible even when no agent is online
Zero Monthly Fee Self-hosted Chatwoot + QwenPaw, no SaaS subscription

Routing Matrix

Inbox ID Site AI Agent Language Role
1 greatqiu.cn sourcing-agent EN Global Sourcing Advisor
7 shopqiu.com halo-blog-agent 简体中文 安防弱电技术顾问

Quick Start

Prerequisites

  • Python 3.8+
  • Self-hosted Chatwoot instance
  • QwenPaw (or any OpenAI-compatible API)

Installation

git clone https://github.com/hanmolabiqiu/chatwoot-ai-agent.git
cd chatwoot-ai-agent
pip install -r requirements.txt

Configuration

Copy and edit the environment file:

cp .env.example .env

Configure your Chatwoot API credentials and AI model settings in .env.

Run

python3 chatwoot_ws_agent.py

The agent will connect to Chatwoot via WebSocket and start listening for messages.

Project Structure

chatwoot-ai-agent/
├── chatwoot_ws_agent.py     # Main WS agent — the core engine
├── knowledge-base.md         # Knowledge base for halo-blog-agent
├── SOUL-halo-blog-agent.md   # AI personality for 安防弱电 (Chinese)
├── .env.example              # Environment configuration template
├── requirements.txt          # Python dependencies
├── README.md                 # This file
└── .gitignore                # Git ignore rules

Version History

Version Date Highlights
v1.0 2026-05-31 Initial release — dual inbox routing, AI↔Human handoff, knowledge base

Roadmap

  • Config Center — Dynamic INBOX_CONFIG from database instead of hardcoded
  • Multi-Tenant Platform — Auto-provision inboxes on payment, user self-service dashboard
  • Amazon API Integration — Product research, order management, customer message sync
  • TikTok Channel — Social commerce integration
  • Tenant Knowledge Base UI — Self-managed KB per tenant
  • Seat-Based Billing — Pay-per-agent pricing model

License

MIT

S
Description
Chatwoot AI Agent
Readme AGPL-3.0 179 KiB
Languages
Python 52.2%
PHP 41.3%
JavaScript 4%
CSS 1.3%
Shell 1.2%