1. Project Overview

Short intro:
Ye project ek multi-agent AI system hai jo accounting ke liye design kiya gaya hai.

Role of agents:

Researcher Agent → DB + local files scan karta hai aur notes banata hai

Planner Agent → Sub-questions aur plan generate karta hai

Summarizer Agent → Final concise structured answer banata hai

Orchestration:
Custom orchestration via run_graph() (LangGraph style), Node.js + Python integrated

2. System Design

Architecture Flow:

User (POST /run-task via Node server)
↓
Python main.py
↓
Orchestrator (run_graph)
↓
[Researcher] → Tools (FileReader, LocalLLM, DB adapter, Calculator)
↓
[Planner] → Breaks into steps
↓
[Summarizer] → Creates final report
↓
Return JSON / Console Output

Agents:

Researcher → Context collection (DB, files, LLM)

Planner → Logical breakdown

Summarizer → Evidence-based final summary

Tools:

Calculator → Basic arithmetic

FileReader → Local text/document scanning

LocalLLM (Ollama) → Offline AI model (no API)

DB Adapters (switchable):

JSON (default)

SQLite

Postgres (stub, extendable)

🔹 3. Setup Instructions

Requirements:

Node.js (≥18)

Python (3.10 recommended, with venv)

Ollama
installed with model (e.g., llama3.2:3b)

Installation Steps (Windows-friendly):

clone repo

git clone https://github.com/trivedijigneshkumar11192/readytensor-multi_agent_module2-Public-.git
cd readytensor-multi_agent_module2-Public-

install Node deps

npm install

setup Python venv

py -3.10 -m venv venv
venv\Scripts\activate
pip install -r python/requirements.txt

run local server

npm start

Test Command:

curl -X POST http://localhost:8080/run-task -H "Content-Type: application/json" -d "{"task":"AI in accounting overview"}"

🔹 4. Sample Input/Output

Input:

Task: AI in accounting overview

Output:

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

🔹 5. Best Practices + Notes

Secure .env files (OLLAMA_PATH, DB configs) → add .env_example

No external APIs → fully offline, reproducible

Modular: new agents or tools can be plugged in easily

Covers ReadyTensor Module 2 requirements (≥3 agents, ≥3 tools, orchestration, DB adapters)