Live Demo: support.datawebify.com
API Docs: support.datawebify.com/docs
Project Page: datawebify.com/projects/agai3_ai_support_agent
Portfolio: datawebify.com | Project 3 of 50
A 5-person support team handling 10,000 tickets per month costs $8,000–$15,000 monthly in salaries alone. Response times average 4–8 hours. Simple, repetitive tickets consume the same human attention as complex, high-value ones. There is no system to route the right ticket to the right handler automatically.
A fully autonomous, multi-agent AI system that classifies every incoming support ticket, auto-resolves simple cases with personalized AI responses, and escalates complex or sensitive cases to human agents with full structured context. The entire pipeline runs in under 30 seconds per ticket.
| Metric | Before | After | Change |
|---|---|---|---|
| Monthly support cost (10K tickets) | $12,000 | $2,800 | -77% |
| Average response time | 4–8 hours | Under 30 seconds | -99% |
| Tickets requiring human agents | 100% | 30–40% | -65% |
| Agent hours consumed per week | 250+ hours | 60–80 hours | -70% |
| Cost per ticket | $1.20 | $0.08–$0.36 | -80% |
Target auto-resolution rate: 60–70% of all Tier-1 tickets
Engagement value: $15,000–$40,000 per deployment
┌─────────────────────────────────────────────┐
│ Orchestrator Agent │
│ (LangGraph — controls flow) │
└──────┬──────────┬──────────┬────────────────┘
│ │ │
▼ ▼ ▼
Classification Response Escalation
Agent Agent Agent
(category, (generates (routes to
urgency, AI reply) human +
complexity) context)
│ │ │
└──────────┴──────────┘
│
▼
Metrics Agent
(tracks resolution,
response time, CSAT)
│
▼
Export Layer
(Supabase + REST API)
Classification Agent Receives each ticket and outputs: category, urgency score (1–5), complexity label (simple or complex), and a confidence score. Powered by GPT-4o-mini with structured JSON output.
Response Agent Generates a personalized, context-aware reply for every auto-resolvable ticket. Pulls customer history from Supabase to avoid generic responses.
Escalation Agent Routes low-confidence or high-complexity tickets to human agents. Attaches a structured context summary so the human never starts from scratch.
Metrics Agent Tracks auto-resolution rate, average response time, escalation rate per category, and cost per ticket in real time.
Export Layer Persists every ticket and outcome to Supabase (PostgreSQL). Supports CSV export for reporting.
| Layer | Technology |
|---|---|
| Language | Python 3.12 |
| Agent Framework | LangGraph |
| AI Model | OpenAI GPT-4o-mini |
| API Layer | FastAPI + Uvicorn |
| Database | Supabase (PostgreSQL) |
| Deployment | Docker + Railway |
| Data Validation | Pydantic v2 |
| HTTP Client | httpx |
| Method | Endpoint | Description |
|---|---|---|
| POST | /ticket |
Submit a new support ticket |
| GET | /tickets |
Retrieve all tickets |
| GET | /tickets/{id} |
Retrieve a single ticket |
| GET | /metrics |
Live business metrics |
| GET | /export/csv |
Download ticket data as CSV |
| GET | /health |
System health check |
Full interactive docs: support.datawebify.com/docs
# 1. Clone the repository
git clone https://github.com/umair801/ai-support-agent.git
cd ai-support-agent
# 2. Create virtual environment
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# 3. Install dependencies
pip install -r requirements.txt
# 4. Configure environment variables
cp .env.example .env
# Add your OPENAI_API_KEY and Supabase credentials to .env
# 5. Run the API
uvicorn main:app --reload
# 6. Open API docs
# http://localhost:8000/docs
# Production endpoints
# Live system: https://support.datawebify.com
# API docs: https://support.datawebify.com/docs
# Health check: https://support.datawebify.com/health
# Metrics: https://support.datawebify.com/metrics
# CSV export: https://support.datawebify.com/export/csv
# Project page: https://datawebify.com/projects/agai3_ai_support_agentPOST https://support.datawebify.com/ticket
{
"customer_id": "cust_001",
"customer_name": "Sarah Mitchell",
"email": "sarah@example.com",
"subject": "Cannot access my account",
"body": "I have been locked out of my account for two days.
I tried resetting my password but never received the email.",
"channel": "email"
}Response (auto-resolved in under 30 seconds):
{
"ticket_id": "tkt_20240315_001",
"status": "auto_resolved",
"classification": {
"category": "account_access",
"urgency": 4,
"complexity": "simple",
"confidence": 0.94
},
"response": "Hi Sarah, I have located your account and triggered a
fresh password reset email...",
"response_time_sec": 8.3,
"metrics": {
"auto_resolution_rate": 0.67,
"avg_response_time_sec": 11.2
}
}AgAI_3_AI_Support_Agent/
├── agents/
│ ├── classification_agent.py
│ ├── response_agent.py
│ ├── escalation_agent.py
│ └── metrics_agent.py
├── graph/
│ └── orchestrator.py
├── models/
│ └── ticket_models.py
├── export/
│ └── supabase_export.py
├── config/
│ └── ticket_config.py
├── main.py
├── metrics_report.py
├── requirements.txt
├── Dockerfile
├── docker-compose.yml
├── .env.example
└── README.md
The system is containerized with Docker and deployed on Railway with a custom domain. Zero-downtime redeploys are handled automatically via Railway's Git integration.
Live system: support.datawebify.com
| Project | Description | Live |
|---|---|---|
| Enterprise WhatsApp Automation | Autonomous WhatsApp outreach and response agent | whatsapp.datawebify.com |
| B2B Lead Generation System | Multi-source lead enrichment and scoring agent | leads.datawebify.com |
| Enterprise AI Support Agent | This project | support.datawebify.com |
Datawebify builds enterprise-grade Agentic AI systems for businesses handling large-scale operations, support workflows, and data pipelines. Each system is production-ready, fully documented, and built to deliver measurable ROI from day one.
Website: datawebify.com
Project Page: datawebify.com/projects/agai3_ai_support_agent
Live System: support.datawebify.com
API Docs: support.datawebify.com/docs
GitHub: github.com/umair801