TechNova Multi‑Agent CRM System

This project is a fully automated stateful multi‑agent CRM platform built using LangGraph, LangChain, FastAPI, SMTP/IMAP, and Google Calendar API.
A custom Streamlit dashboard provides an interactive UI for monitoring agents, approving actions, and demonstrating the workflow.

---

🚀 Project Overview

TechNova is a fictional AI‑driven consulting company that helps B2B clients modernize operations using AI, cloud, and MLOps.
This CRM system automates the entire customer acquisition pipeline:

1. Lead Discovery (Recruitment Agent)

   • Analyses synthetic dataset of potential customers.
   • Assigns intent scores and shortlists top clients.

2. Email Outreach (Interaction Agent)

   • Generates personalized email drafts using LLMs.
   • Allows user to edit & approve emails in UI.
   • Sends emails via Gmail SMTP.
   • Continuously polls inbox via IMAP for replies.

3. Scheduling (Scheduler Agent)

   • Extracts meeting availability from customer reply.
   • Auto‑schedules meetings in Google Calendar.
   • Sends real calendar invites.

4. Call Analytics (Analytics Agent)

   • Ingests call transcript (demo uses synthetic one).
   • Produces summary, key themes, pain points, next best action.
   • Stores results in summary file.

5. Supervisor Agent

   • Orchestrates full stateful workflow end‑to‑end.

6. Streamlit UI

   • Shows each agent step clearly.
   • Users manually approve key actions (email sending, scheduling).
   • Real‑time inbox monitoring + activity logs.

---

🛠️ Setup Instructions

1. Install Dependencies

pip install -r requirements.txt

Minimum required libraries:

streamlit
langchain
langgraph
langchain-openai
google-api-python-client
python-dotenv

2. Configure Environment Variables

Create .env file:

SMTP_EMAIL=your-email@gmail.com
SMTP_PASS=your-app-password
IMAP_EMAIL=your-email@gmail.com
IMAP_PASS=your-app-password
OPENAI_API_KEY=your-openai-key

3. Google Calendar API Setup

• Go to Google Cloud Console
• Enable Calendar API
• Download OAuth credentials.json
• Place it in your project root
• First run will open browser to authenticate
• Generates token.json

---

▶️ Running the App

Start Streamlit UI

streamlit run streamlit_app.py

Run Supervisor Agent only

python main.py

---

🤖 How It Works (Workflow Summary)

Recruitment Agent

• Reads synthetic customer dataset
• Uses LLM to assign intent scores
• Produces ranked shortlist

Interaction Agent

• Generates email drafts (editable in UI)
• Sends approved emails via SMTP
• Uses IMAP polling to detect replies

Scheduler Agent

• Parses reply sentiment & availability
• Schedules meeting in Google Calendar
• Sends invite to customer

Analytics Agent

• Takes call transcript (synthetic for demo)
• Extracts summary, key insights, action items
• Saves results in summary.txt

---

🎯 Features

Feature	Description
🔍 Lead Analysis	AI ranking of client intent
✉️ Email Automation	SMTP + LLM email drafts
📬 Auto Reply Detection	IMAP inbox scanning
📅 Meeting Scheduling	Google Calendar integration
📊 Call Analysis	LLM-powered transcript analytics
📟 Streamlit UI	Real‑time agent monitoring

---

📦 Deliverables

• Fully functional streamlit dashboard
• End‑to‑end multi‑agent CRM pipeline
• Real email + calendar scheduling
• Editable outreach drafts
• Stateful orchestration logic
• Synthetic datasets for lead discovery & transcripts (Replace it with real data, use Crunchbase or LinkedIn APIs )

---

🏆 This Project Demonstrates

✔ Multi-agent orchestration with LangGraph
✔ AI‑driven customer acquisition
✔ Real email + calendar tool usage
✔ Generative AI for outreach & analytics
✔ Event-driven agent triggers
✔ Production-like CRM workflow in Python

---