Enterprise AI Governance, Safety & Observability Platform
Monitor. Govern. Trust your AI.
GuardForge is a full-stack platform for governing, monitoring, and securing enterprise AI agents. It provides a centralized command center where teams can enforce safety policies, trace agent decisions in real-time, run red-team simulations, and generate compliance reports — all through a stunning, modern dashboard.
Whether you're deploying customer support bots, billing agents, or multi-agent swarms, GuardForge gives you the visibility and control to ensure every AI interaction is safe, transparent, and compliant.
| Feature | Description |
|---|---|
| 📊 Live Dashboard | Real-time metrics for agents, workflows, violations, and risk scores with interactive charts |
| 🔍 Trace Explorer | WebSocket-powered live execution traces showing step-by-step agent decision-making |
| 📜 AI Policy Generator | Upload compliance documents (PDF/Text) and auto-generate governance rules |
| 🤖 Agent Architect | Visual builder for configuring AI agents with custom guardrails and tools |
| 💓 Agent Health Monitor | Track latency, faithfulness scores, and operational status across all agents |
| 🐝 Swarm Intelligence | Orchestrate and monitor multi-agent collaborative tasks |
| 🔴 Red Team Simulations | Automated adversarial testing with prompt injection, jailbreak, and PII leak scenarios |
| 🔒 Safety Scanner | Comprehensive vulnerability scanning across agent configurations |
| 📋 Audit Logs | Immutable, searchable trail of every system event with severity classification |
| 👤 User Profiles | Secure authentication with role-based access control (Admin / Analyst / Viewer) |
| 📄 Compliance Reports | One-click PDF report generation for regulatory audits |
GuardForge/
├── frontend/ # React + Vite SPA
│ ├── src/
│ │ ├── api/ # Axios client with JWT interceptor
│ │ ├── components/ # Reusable UI components (Layout, Card)
│ │ ├── context/ # AuthContext for session management
│ │ ├── hooks/ # Custom hooks (useWebSocket)
│ │ └── pages/ # 17 feature pages
│ └── package.json
│
├── backend/ # FastAPI application
│ ├── app/
│ │ ├── api/ # Route handlers (REST + WebSocket)
│ │ ├── core/ # Config, security utilities, swarm engine
│ │ ├── dependencies/ # Auth middleware (JWT + RBAC)
│ │ ├── models/ # SQLAlchemy ORM models
│ │ └── database.py # DB session management
│ └── requirements.txt
│
├── docker-compose.yml # PostgreSQL + Redis + App stack
├── package.json # Root orchestration scripts
└── .env # Environment configuration
- Node.js ≥ 18
- Python ≥ 3.10
- npm ≥ 9
git clone https://github.com/Aliyan-cmd/GuardForge.git
cd GuardForge# Install frontend dependencies
npm install --prefix frontend
# Create and activate Python virtual environment
python -m venv backend/venv
source backend/venv/bin/activate # Linux/macOS
# backend\venv\Scripts\activate # Windows
# Install backend dependencies
pip install -r backend/requirements.txtexport PYTHONPATH=$PYTHONPATH:$(pwd)/backend
python scratch/seed_admin.pyThis creates all database tables and seeds a default admin account.
npm run devThis starts both servers concurrently:
| Service | URL |
|---|---|
| Frontend | http://localhost:5173 |
| Backend API | http://localhost:8000 |
| API Docs | http://localhost:8000/docs |
| Field | Value |
|---|---|
admin@guardforge.ai |
|
| Password | admin123 |
For a production-like environment with PostgreSQL and Redis:
# Update .env to use PostgreSQL
# DATABASE_URL=postgresql://guardforge:secret@db:5432/guardforge
npm run docker-upTo tear down:
npm run docker-downGuardForge uses JWT-based authentication with role-based access control:
| Role | Permissions |
|---|---|
| Admin | Full access — manage users, create policies, run red-team tests, generate reports |
| Analyst | View dashboards, run workflows, manage agents, search audit logs |
| Viewer | Read-only access to dashboards and analytics |
Login → JWT Issued → Stored in localStorage → Sent via Authorization: Bearer <token>
| Technology | Purpose |
|---|---|
| React 18 | Component-based UI framework |
| Vite 8 | Lightning-fast build tool and dev server |
| React Router v6 | Client-side routing with v7 future flags |
| Lucide React | Premium icon library |
| Axios | HTTP client with JWT interceptor |
| WebSocket API | Real-time trace streaming |
| Technology | Purpose |
|---|---|
| FastAPI | High-performance async Python API framework |
| SQLAlchemy 2.0 | ORM for database operations |
| SQLite / PostgreSQL | Relational database (dev / prod) |
| python-jose | JWT token creation and validation |
| passlib + bcrypt | Secure password hashing |
| PyPDF2 | PDF document parsing for policy extraction |
| reportlab | Dynamic PDF compliance report generation |
| uvicorn | ASGI server |
| Model | Description |
|---|---|
User |
Credentials, hashed passwords, roles (admin, analyst, viewer) |
Agent |
Registered AI entities with status tracking |
Policy |
Versioned governance rules with JSON rule definitions |
PolicyVersion |
Historical snapshots of policy changes |
Workflow |
Ordered sequences of agents for business processes |
ExecutionLog |
Records of workflow runs and outcomes |
AuditLog |
Immutable event trail with severity levels |
GuardrailViolation |
Captured policy violations with context |
ApprovalRequest |
Human-in-the-loop approval workflows |
RedTeamTest |
Adversarial test configurations and results |
| Method | Endpoint | Description |
|---|---|---|
POST |
/auth/login |
Authenticate and receive JWT |
POST |
/auth/signup |
Register a new user account |
| Method | Endpoint | Description |
|---|---|---|
GET |
/dashboard/overview |
Real-time platform metrics |
POST |
/dashboard/seed-demo |
Initialize demo environment |
| Method | Endpoint | Description |
|---|---|---|
GET |
/agents/ |
List all registered agents |
| Method | Endpoint | Description |
|---|---|---|
GET |
/policies/ |
List all governance policies |
POST |
/policies/ |
Create a new policy |
POST |
/policies/generate-from-doc |
AI-powered policy generation from documents |
| Method | Endpoint | Description |
|---|---|---|
GET |
/workflows/ |
List all workflows |
POST |
/workflows/run/{id} |
Execute a workflow |
WS |
/workflows/ws/run/{run_id} |
Live execution trace stream |
| Method | Endpoint | Description |
|---|---|---|
GET |
/audit/logs |
Retrieve audit log entries |
GET |
/audit/search?q= |
Natural language log search |
| Method | Endpoint | Description |
|---|---|---|
GET |
/reports/compliance/{id} |
Generate PDF compliance report |
# Run full stack (frontend + backend)
npm run dev
# Run frontend only
npm run dev-frontend
# Run backend only
npm run dev-backend
# Install all dependencies
npm run install-all
# Docker
npm run docker-up
npm run docker-down- Real LLM integration (OpenAI / LangChain) for policy generation
- Full red-team automation with adversarial agent framework
- PostgreSQL migration for production persistence
- LDAP / SSO integration for enterprise auth
- Webhook notifications for critical violations
- Kubernetes deployment manifests
- Multi-tenant organization support
- Export audit logs to SIEM platforms
Contributions are welcome! Please follow these steps:
- Fork the repository
- Create your feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
This project is licensed under the MIT License. See the LICENSE file for details.
Aliyan
GitHub: @Aliyan-cmd
Built with ❤️ for safer AI
If you find this project useful, consider giving it a ⭐