🚀 Production-Ready Multi-Agent AI System for Intelligent Repository Analysis & Automation

A fully modular, production-ready multi-agent AI system designed to intelligently analyze GitHub repositories, extract metadata, summarize content, and improve documentation.

This project represents the final capstone (Project 3) of the Agentic AI Developer Certification and is an upgraded, production-grade version of the previous Project 2: Multi-Agent System.

📌 About This Project
This system transforms a functional prototype into a reliable, secure, and user-ready AI application, featuring:

🧠 Multi-agent collaboration
🛡️ Guardrails & safety validation
🧪 Comprehensive testing (unit + integration + E2E)
🖥️ Streamlit user interface
🔧 Robust tooling for repo parsing & metadata extraction
📈 Production-readiness enhancements

🧩 System Architecture Overview
The system uses a network of coordinated AI agents, each specializing in a specific task:
🤖 Agents
Agent Role

• AnalyzerAgent Generates structured summaries of repository content
• MetadataAgent Extracts tech stack, dependencies, and project intent
• ImproverAgent Suggests documentation and structural improvements

🛠 Tools
Tool Function

• Repo Reader Reads README and file tree from a local repo
• Web Search Tool Performs DuckDuckGo search for context (with fallback)
• Guardrails System Input validation + output filtering + safety checks
• Logger Utility Tracks internal operations & errors

🖥️ User Interface
A lightweight Streamlit application enables:

• Uploading / entering a repository path
• Running analysis with a single click
• Viewing AI-generated summaries, metadata, and suggestions

🏗 Folder Structure
multi-agent-system/
│
├── agents/
│ ├── analyzer_agent.py
│ ├── metadata_agent.py
│ ├── improver_agent.py
│
├── tools/
│ ├── repo_reader.py
│ ├── web_search_tool.py
│
├── utils/
│ ├── guardrails.py
│ ├── logger.py
│
├── ui/
│ ├── app.py
│
├── tests/
│ ├── test_analyzer_agent.py
│ ├── test_metadata_agent.py
│ ├── test_improver_agent.py
│ ├── test_end_to_end.py
│
└── main.py

🧠 How It Works
The system follows a deterministic, multi-agent orchestration pipeline:
1️⃣ Repo Reader collects repository content
2️⃣ AnalyzerAgent generates a structured summary
3️⃣ MetadataAgent extracts tech stack & core metadata
4️⃣ ImproverAgent proposes documentation & quality improvements
5️⃣ All outputs merge in the orchestrator
6️⃣ Final results displayed in Streamlit UI

🧬 Architecture Diagram
flowchart TD

A[User Input
Repository Path] --> B[Repo Reader]

B --> C[AnalyzerAgent]
B --> D[MetadataAgent]
B --> E[ImproverAgent]

C --> F[Summary Output]
D --> G[Metadata Output]
E --> H[Improvement Suggestions]

F --> I[Workflow Orchestrator]
G --> I
H --> I

I --> J[Streamlit UI
Final JSON Output]

🛡 Safety, Guardrails & Fallbacks
To ensure a production-grade experience, the system includes:

• Input sanitization
• Minimal-length validation
• Output safety filtering
• Anti-prompt-injection checks
• Automatic LLM failure fallback (keyword-based analysis)
• Deterministic DummyLLM for tests

🧪 Testing & Reliability
The testing suite ensures functional accuracy and system stability:

• Test Type Purpose
• Unit Tests Validate each agent and tool
• Integration Tests Validate multi-agent communication
• E2E Test Confirms the full workflow runs successfully
• DummyLLM Ensures predictable and deterministic tests

Run all tests:
pytest -v

🖥️ UI: Streamlit Application
Launch the UI:

cd ui
streamlit run app.py

The interface displays:

• Summary
• Metadata
• Suggested improvements
  in a structured JSON format.

⚙️ Installation
1️⃣ Clone the repository
git clone https://github.com/pranteshdahikar/multi-agent-system
cd multi-agent-system

2️⃣ Install dependencies
pip install -r requirements.txt

3️⃣ Run the analyzer workflow
python main.py

📤 Example Output
{
"analysis_summary": "This repository implements a production-ready multi-agent system...",
"metadata": {
"tech_stack": ["Python", "Streamlit", "LangChain"],
"project_purpose": "Automated repository analysis & documentation enhancement"
},
"improvements": [
"Add architecture diagram",
"Add contribution guidelines",
"Improve README structure"
]
}

🔗 Related GitHub Projects
These projects support this certification journey:
Project Links:
Project – RAG Assistant https://github.com/pranteshdahikar/rag-assistant
Project – Multi-Agent System https://github.com/pranteshdahikar/multi-agent-system