A Production-Ready Multi-Agent AI System powered by Gemini
👉 Live App: https://agentforge-ai-100704.streamlit.app
[AgentForge-AI UI]
AgentForge-AI is a modular, production-oriented multi-agent AI system designed to solve complex tasks through structured collaboration between specialized AI agents.
Instead of a single monolithic LLM call, the system decomposes a task into planning, research, implementation, and review, closely mirroring how real-world engineering teams operate.
This project demonstrates:
- Multi-agent orchestration
- Clean system design
- LLM integration (Google Gemini)
- Streamlit-based UI
- Testability and extensibility
- Deployment readiness
Modern AI applications often fail due to:
- Unstructured prompts
- Lack of reasoning separation
- No quality verification
- Poor scalability
AgentForge-AI solves these problems by design.
- 🧠 Planner Agent — breaks down tasks logically
- 🔍 Researcher Agent — gathers best practices and tools
- 💻 Coder Agent — produces implementation-level output
- 🧪 Reviewer Agent — evaluates quality and readiness
- 🧩 Orchestrator — manages agent execution flow
- 🧠 Memory Layer — enables extensibility for future context handling
This architecture enables transparent reasoning, higher-quality outputs, and enterprise-ready extensibility.
User Task
↓
Planner Agent
→ Task Decomposition
↓
Researcher Agent
→ Concepts & Tools
↓
Coder Agent
→ Implementation / Code
↓
Reviewer Agent
→ Quality & Readiness Evaluation
Each agent is:
- Independent
- Prompt-specialized
- Replaceable
- Testable
AgentForge-AI/
│
├── app.py # Streamlit UI entry point
├── main.py # CLI / backend execution
├── requirements.txt # Project dependencies
├── Dockerfile # Containerized deployment
├── .env # Environment variables (API keys)
├── .gitignore
│
├── agents/
│ ├── __init__.py
│ ├── base.py # Abstract base agent
│ ├── planner.py # Task planning agent
│ ├── researcher.py # Research & analysis agent
│ ├── coder.py # Code generation agent
│ └── reviewer.py # Evaluation & scoring agent
│
├── core/
│ ├── __init__.py
│ ├── gemini_client.py # Gemini API wrapper
│ ├── orchestrator.py # Agent execution pipeline
│ └── memory.py # Memory abstraction (extensible)
│
├── tests/
│ ├── test_planner.py
│ ├── test_researcher.py
│ ├── test_coder.py
│ ├── test_reviewer.py
│ └── test_orchestrator.py
│
└── README.md
Design a REST API for student course enrollment
- Structured API design
- Resource modeling
- Endpoints and HTTP methods
- Error handling strategy
- Production-grade FastAPI implementation
- Quality and readiness scoring
This mirrors how real backend systems are architected in industry.
- Python 3.10+
- Google Gemini 2.5 Flash
- Streamlit
- Pydantic
- FastAPI (generated output)
- python-dotenv
- pytest
- Docker
git clone https://github.com/rajveer100704/AgentForge-AI.git
cd AgentForge-AI
python -m venv venv
Activate:
Windows
venv\Scripts\activate
Linux / macOS
source venv/bin/activate
pip install -r requirements.txt
Create a .env file in the project root:
GEMINI_API_KEY=your_gemini_api_key_here
.env to GitHub.
streamlit run app.py
Then open:
http://localhost:8501
pytest
All agents and orchestration logic are independently testable.
You can run AgentForge-AI using Docker without installing Python locally.
docker build -t agentforge-ai .
docker run -p 8501:8501 \
-e GEMINI_API_KEY=your_api_key_here \
agentforge-ai
Then open:
http://localhost:8501
- Persistent long-term memory (Vector DB)
- Agent self-reflection loops
- Role-based access control
- Plug-and-play agent registry
- Async / parallel agent execution
- Execution analytics dashboard
- CI/CD pipeline with GitHub Actions
Rajveer Saggu
📧 Email: rajveer19255@gmail.com
🌐 GitHub: https://github.com/rajveer100704
If you like this project, give it a ⭐ —
it boosts visibility and motivates future development.