Multi-Agent Clientele Services

Enterprise Multi-Agent Customer Service Platform built with Next.js 15, LangChain/LangGraph, and modern AI infrastructure.

Architecture

User → Next.js (PWA) → Socket.IO → LangGraph Agent Pipeline
                                    ├── Intent Classifier
                                    ├── Presale Agent (RAG + Vector Search)
                                    ├── AfterSale Agent (RAG + Complaint Matching)
                                    ├── CV Agent (Image Analysis)
                                    └── Supervisor Agent (Escalation)
                                         │
                                    ┌────┴────┐
                                    │  PostgreSQL + pgvector
                                    │  Redis (Cache / Session)
                                    │  Milvus (Vector DB)
                                    └─────────┘

Tech Stack

Layer	Technology
Frontend	Next.js 15, React 19, Tailwind CSS, Framer Motion
Agent Framework	LangChain 0.3 + LangGraph 0.2
Database	PostgreSQL 16 + pgvector
Vector Search	Milvus 2.4 (primary), pgvector (fallback)
Cache	Redis 7
Real-time	Socket.IO 4.8
ORM	Prisma 6
CV Model	Adapter pattern (OpenAI / Claude / local)
LLM	DeepSeek (OpenAI-compatible)
Embedding	BGE-M3 (Ollama, 1024-dim)

Quick Start

Prerequisites

• Docker Desktop
• Node.js 18+
• PowerShell (for setup scripts)

Setup

# 1. Install dependencies
npm install

# 2. Configure environment
cp .env.example .env.local
# Edit .env.local with your API keys

# 3. Start infrastructure (PostgreSQL + Redis + Milvus)
docker compose up -d

# 4. Initialize database
npx prisma generate
npx prisma db push
npm run prisma:seed

# 5. Start development
npm run dev       # Next.js dev server (port 3000)
npm run server    # Socket.IO server

Or use the all-in-one script:

npm run quick

Setup Scripts

npm run setup           # Full automated setup
npm run setup:bge-m3    # BGE-M3 embedding model setup

Project Structure

src/
├── app/          # Next.js App Router pages & API routes
├── components/   # React components (chat, dashboard, agents)
├── hooks/        # Custom React hooks
├── lib/          # Utility libraries & agent logic
└── types/        # TypeScript type definitions

server/           # Socket.IO server
prisma/           # Database schema & migrations
mcp/              # MCP (Model Context Protocol) extensions
docs/             # Documentation

Features

• Multi-Agent Pipeline: Intent classification → agent routing → RAG response generation
• Real-time Chat: Socket.IO-based messaging with typing indicators
• Vector Search: Semantic FAQ matching via Milvus/pgvector + BGE-M3 embeddings
• CV Analysis: Multi-model image analysis (product defect, screenshot OCR)
• Multi-Tenant: Row-Level Security with tenant isolation
• PWA: Installable, offline-capable progressive web app
• MCP Extensions: Model Context Protocol for agent tool exposure

Environment Variables

See .env.example for the full list. Key variables:

Variable	Description
DATABASE_URL	PostgreSQL connection string
REDIS_URL	Redis connection string
MILVUS_ADDRESS	Milvus server address
OPENAI_API_KEY	LLM API key (DeepSeek or OpenAI)
OPENAI_BASE_URL	LLM API endpoint
EMBEDDING_BASE_URL	Embedding model endpoint
CV_MODEL_*	Computer Vision model configuration

License

Private — All rights reserved.

如果这个项目对你有帮助，请给一个 ⭐ Star！