https://github.com/salman-aziz-4425/support-chatbot
A Support Chatbot demonstrating Autogen handoff pattern
https://github.com/salman-aziz-4425/support-chatbot
ai-agents autogen fastapi gen ollama python
Last synced: about 2 months ago
JSON representation
A Support Chatbot demonstrating Autogen handoff pattern
- Host: GitHub
- URL: https://github.com/salman-aziz-4425/support-chatbot
- Owner: salman-aziz-4425
- Created: 2025-07-27T01:43:49.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-07-28T05:14:48.000Z (3 months ago)
- Last Synced: 2025-08-19T23:17:30.096Z (2 months ago)
- Topics: ai-agents, autogen, fastapi, gen, ollama, python
- Language: Python
- Homepage:
- Size: 79.2 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 🤖 Multi-Agent Customer Support System
A sophisticated customer support system with AI agents, human agents, and seamless handoffs between them.
## 📁 Project Structure
```
autogen-project/
├── main.py # Main application entry point
├── requirements.txt # Python dependencies
├── README.md # This file
├── src/ # Source code
│ ├── agents/ # Agent implementations
│ │ ├── __init__.py
│ │ ├── ai_agent.py # AI agent base class
│ │ ├── tools.py # Function tools and delegate tools
│ │ └── websocket_agents.py # WebSocket agents for human/user
│ ├── api/ # API endpoints
│ │ ├── __init__.py
│ │ ├── routes.py # REST API routes
│ │ └── websockets.py # WebSocket endpoints
│ ├── models/ # Data models
│ │ ├── __init__.py
│ │ └── message_models.py # Pydantic models for messages
│ ├── services/ # Business logic services
│ │ ├── __init__.py
│ │ ├── agent_runtime.py # Agent runtime initialization
│ │ ├── connection_manager.py # WebSocket connection management
│ │ └── transfer_service.py # Human-to-AI transfer logic
│ └── utils/ # Utility functions
│ ├── __init__.py
│ └── serialization.py # Message serialization utilities
├── static/ # Static files
│ ├── index.html # Customer chat interface
│ ├── agent_dashboard.html # Human agent dashboard
│ ├── style.css # Stylesheets
│ └── script.js # Frontend JavaScript
└── autogen_env/ # Virtual environment
```
## 🚀 Quick Start
1. **Install dependencies:**
```bash
pip install -r requirements.txt
```
2. **Start the server:**
```bash
python main.py
```
3. **Access the interfaces:**
- Customer Chat: http://localhost:8000
- Agent Dashboard: http://localhost:8000/agent-dashboard
## 🏗️ System Architecture
```mermaid
graph TB
%% Client Layer
subgraph "Client Layer"
Customer["👤 Customer WebApp"]
Agent["👨💼 Human Agent Dashboard"]
end
%% API & WebSocket Layer
subgraph "Communication Layer"
API["🌐 FastAPI Routes"]
WS_Customer["🔗 Customer WebSocket"]
WS_Agent["🔗 Agent WebSocket"]
end
%% Core Agent Runtime
subgraph "AutoGen Agent Runtime"
Runtime["⚡ SingleThreadedAgentRuntime"]
subgraph "AI Agents"
Triage["🎯 Triage Agent"]
Technical["🔧 Technical Agent"]
Billing["💳 Billing Agent"]
Sales["🛒 Sales Agent"]
end
subgraph "WebSocket Agents"
WSHuman["👤 WebSocket Human Agent"]
WSUser["📱 WebSocket User Agent"]
end
end
%% Tools & Services
subgraph "Tools & Services"
RegularTools["🛠️ Regular Tools"]
TransferTools["🔄 Transfer Tools"]
end
%% Data & State Management
subgraph "Services Layer"
ConnMgr["📊 Connection Manager"]
TransferSvc["🔀 Transfer Service"]
LLM["🤖 Ollama LLM Client"]
end
%% Message Flow
Customer --> WS_Customer
Agent --> WS_Agent
Agent --> API
WS_Customer --> WSUser
WS_Agent --> WSHuman
API --> ConnMgr
WSUser --> Runtime
WSHuman --> Runtime
Runtime --> Triage
Runtime --> Technical
Runtime --> Billing
Runtime --> Sales
Triage -.->|transfer_to_technical| Technical
Triage -.->|transfer_to_billing| Billing
Triage -.->|transfer_to_sales| Sales
Triage -.->|escalate_to_human| WSHuman
Technical -.->|transfer_back| Triage
Billing -.->|transfer_back| Triage
Sales -.->|transfer_back| Triage
Technical --> RegularTools
Billing --> RegularTools
Sales --> RegularTools
Triage --> TransferTools
Technical --> TransferTools
Billing --> TransferTools
Sales --> TransferTools
Runtime --> LLM
Runtime --> ConnMgr
WSHuman --> TransferSvc
%% Response Flow
WSUser --> WS_Customer
WSHuman --> WS_Agent
%% Styling
classDef clientLayer fill:#3498db,stroke:#2980b9,stroke-width:2px,color:#fff
classDef commLayer fill:#e74c3c,stroke:#c0392b,stroke-width:2px,color:#fff
classDef agentLayer fill:#f39c12,stroke:#d68910,stroke-width:2px,color:#fff
classDef serviceLayer fill:#27ae60,stroke:#229954,stroke-width:2px,color:#fff
classDef aiAgent fill:#9b59b6,stroke:#8e44ad,stroke-width:2px,color:#fff
classDef wsAgent fill:#1abc9c,stroke:#16a085,stroke-width:2px,color:#fff
class Customer,Agent clientLayer
class API,WS_Customer,WS_Agent commLayer
class Runtime agentLayer
class Triage,Technical,Billing,Sales aiAgent
class WSHuman,WSUser wsAgent
class ConnMgr,TransferSvc,LLM,RegularTools,TransferTools serviceLayer
```
### Component Details
**🎯 Triage Agent**
- Routes customer requests to appropriate specialists
- Tools: Account lookup, transfer functions
- Handles initial customer contact and request routing
**🔧 Technical Support Agent**
- Hardware, software, and system troubleshooting
- Tools: Ticket creation, system status checks
- Provides step-by-step solutions and escalation handling
**💳 Billing Support Agent**
- Payment, subscription, and billing inquiries
- Tools: Account lookup, ticket creation
- Handles payment guidance and refund assistance
**🛒 Sales Support Agent**
- Product information and purchase assistance
- Tools: Account lookup
- Provides feature explanations and pricing information
**👤 Human Support Agent**
- Complex issue resolution and escalation handling
- Full system override with AI agent transfer capabilities
- Dedicated dashboard with real-time customer assignment
## Key Components
### 🎯 **Triage Agent**
- **Role**: Initial customer contact and request routing
- **Tools**: Account lookup, transfer functions
- **System Message**: Routes customers to appropriate specialists
- **Transfers To**: Technical, Billing, Sales, or Human agents
### 🔧 **Technical Support Agent**
- **Role**: Hardware, software, and system troubleshooting
- **Tools**: Ticket creation, system status checks
- **Capabilities**: Step-by-step solutions, escalation handling
- **Transfers**: Back to triage or human escalation
### 💳 **Billing Support Agent**
- **Role**: Payment, subscription, and billing inquiries
- **Tools**: Account lookup, ticket creation
- **Capabilities**: Payment guidance, refund assistance
- **Security**: Recommends human verification for sensitive data
### 🛒 **Sales Support Agent**
- **Role**: Product information and purchase assistance
- **Tools**: Account lookup
- **Capabilities**: Feature explanations, pricing information
- **Approach**: Helpful without being pushy
### 👤 **Human Support Agent**
- **Role**: Complex issue resolution and escalation handling
- **Capabilities**: Full system override, AI agent transfers
- **Tools**: Transfer back to any AI agent with context
- **Interface**: Dedicated dashboard with real-time customer assignment
## Message Flow Architecture
### Customer to AI Flow
```
Customer Input → WebSocket → Context Building → Triage Agent → LLM Processing → Tool/Transfer Decision → Response/Transfer
```
### AI Agent Processing
```mermaid
sequenceDiagram
participant C as Customer
participant WS as WebSocket
participant T as Triage Agent
participant TA as Technical Agent
participant LLM as Ollama LLM
C->>WS: "My laptop won't boot"
WS->>T: UserTask with context
T->>LLM: Process with system message + tools
LLM->>T: transfer_to_technical()
T->>TA: UserTask with transfer context
TA->>LLM: Process technical issue
LLM->>TA: create_support_ticket()
TA->>TA: Execute tool
TA->>LLM: Get response with ticket info
LLM->>TA: Final response
TA->>WS: AgentResponse
WS->>C: Technical solution + ticket ID
```
### Human Escalation Flow
```mermaid
sequenceDiagram
participant C as Customer
participant AI as AI Agent
participant H as Human Agent
participant M as Connection Manager
C->>AI: Complex request
AI->>AI: escalate_to_human()
AI->>M: Check available agents
M->>H: Assign customer + context
H->>C: Direct communication
H->>AI: transfer_to_technical()
AI->>C: Continues with context
```
## Technology Stack
- **Framework**: Microsoft AutoGen Core
- **LLM**: Ollama (llama3.1:latest)
- **Backend**: FastAPI + WebSockets
- **State Management**: In-memory with Connection Manager
- **Agent Types**: RoutedAgent with message handlers
- **Tools**: Function-based with JSON schema
- **Communication**: Real-time bidirectional WebSocket
## Key Features
- ✅ **Intelligent Routing**: Context-aware customer request routing
- ✅ **Stateful Conversations**: Full context preservation across transfers
- ✅ **Seamless Escalation**: Human agents can take over any conversation
- ✅ **Bidirectional Transfers**: Humans can transfer back to AI specialists
- ✅ **Real-time Communication**: WebSocket-based instant messaging
- ✅ **Agent Specialization**: Domain-specific tools and prompts
- ✅ **Error Handling**: Graceful degradation and retry mechanisms
- ✅ **Monitoring Dashboard**: Real-time system status and agent availability
## Agent Capabilities Matrix
| Agent Type | Account Lookup | Ticket Creation | Status Check | Transfer | Human Escalation |
|------------|---------------|----------------|--------------|----------|------------------|
| Triage | ✅ | ❌ | ❌ | ✅ | ✅ |
| Technical | ❌ | ✅ | ✅ | ✅ | ✅ |
| Billing | ✅ | ✅ | ❌ | ✅ | ✅ |
| Sales | ✅ | ❌ | ❌ | ✅ | ✅ |
| Human | ✅ | ✅ | ✅ | ✅ | N/A |