1. Introduction & Context

This project is the final capstone of the Agentic AI Developer Certification (AAIDC). It extends the prototype from Module 2 into a production-ready, safe, tested, and user-friendly system.

Agentic AI systems are powerful, but production deployment requires additional features such as guardrails, testing, monitoring, and a user-friendly interface. This project demonstrates how to take a multi-agent system into real-world readiness.

2. Architecture Overview

A. Agents:

Researcher Agent → Collects context from DB, local files, and LLM.

Planner Agent → Breaks down complex tasks into smaller logical steps.

Summarizer Agent → Creates a concise, structured final answer.

B. Tools:

Calculator → Arithmetic & basic math operations.

File Reader → Reads and extracts text from local files.

Local LLM → Offline Ollama-based LLM integration.

DB Adapters → JSON (default), SQLite, Postgres (extendable).

C. Orchestration:

Custom orchestration built via simple_graph.py.

Each agent is a node in the execution graph, orchestrated for safe, sequential workflow.

📌 Architecture Diagram (use this PNG):

3. Features

✅ 3+ agents with orchestration.
✅ 3+ tools integrated.
✅ Safety guardrails (input/output sanitization, fallback).
✅ Streamlit web UI for user-friendly interaction.
✅ Node.js server wrapper for backend integration.
✅ Logging & UTF-8 safe output.
✅ Pytest-based test suite with ~70% coverage.

4. Installation

A. Clone Repository

git clone https://github.com/trivedijigneshkumar11192/multi_agent_advance_module3.git
cd multi_agent_advance_module3

B. Setup Virtual Environment (Windows)

python -m venv venv
venv\Scripts\activate

C. Install Dependencies

// Python
pip install -r python/requirements.txt

// Node.js backend
cd backend
npm install

D. Running the Project

Start Backend

npm start

Run Directly

python python/main.py --task "AI in accounting overview" --dbtype sqlite

Run Web UI

streamlit run python/ui.py

E. Testing

Run the test suite with coverage reports:

pytest --cov --cov-report=term-missing --cov-report=html

5. Sample Input/Output

A. Input:

{"task": "AI in accounting overview", "dbtype": "sqlite"}

B. Output:

{
"ok": true,
"output": "Task: AI in accounting overview\n\n== Plan == ...\n== Evidence == ...\n== Summary == ..."
}

6. Future Work

Add more specialized agents (e.g., Compliance Agent, Finance Analyzer).

Advanced evaluation metrics for agent workflows.

Support for additional UI integrations.

CI/CD pipeline with Docker & deployment recipes.

License:

This project is released under the MIT License.
It is intended for research and educational use, showcasing a modular and production-ready multi-agent system architecture.