AI Communication Assistant (WhatsApp + Gmail)

An end-to-end AI automation system that reads, understands, and replies to messages from:

• WhatsApp (via webhook API)
• Gmail (via Gmail API)

Powered by a locally running Qwen LLM, this system generates intelligent, human-like replies without relying on external paid APIs.

---

Overview

This project demonstrates how to build a real-world AI assistant that integrates with:

• WhatsApp (via Meta Cloud API + Webhooks)
• Gmail (via Gmail API + OAuth)
• Local LLM (Qwen via Hugging Face)

Unlike typical AI apps, this runs fully locally without paid APIs.

---

KEY Features

• Real-time WhatsApp auto-replies
• Automated Gmail responses
• Multi-user session memory
• Local LLM inference (Qwen 0.5B)
• Email cleaning & truncation
• Secure environment variable handling

---

Architecture

WhatsApp Flow (Webhook-based)

┌──────────────┐
│ WhatsApp API │
└──────┬───────┘
       │  (Webhook via ngrok)
       ▼
┌──────────────┐
│  FastAPI App │   ← main.py
└──────┬───────┘
       ▼
┌──────────────┐
│  LLM Model   │   ← model.py
└──────────────┘

---

Gmail Flow (Independent Script)

┌──────────────┐
│  Gmail API   │
└──────┬───────┘
       ▼
┌──────────────┐
│ Gmail Script │   ← gmail.py
└──────┬───────┘
       ▼
┌──────────────┐
│  LLM Model   │   ← model.py
└──────────────┘

---

Setup Guide

1. Clone Repo

git clone https://github.com/YOUR_USERNAME/YOUR_REPO.git
cd YOUR_REPO 

2. Environment Setup

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

3. Configure Environment Variables

Create .env:

EMAIL=your_email@gmail.com
OAUTH_CREDENTIALS_FILE=Oauth_cred_chatbot_gmail.json
VERIFY_TOKEN=your_verify_token
ACCESS_TOKEN=your_whatsapp_access_token
PHONE_NUMBER_ID=your_phone_number_id

---

API Setup

1. WhatsApp (Meta Cloud API) -Create app → Add WhatsApp Get ACCESS_TOKEN, PHONE_NUMBER_ID -Set webhook: https://your-ngrok-url/webhook

• ngrok

ngrok http 8000

2. Gmail API Enable Gmail API in Google Cloud Setup OAuth Consent Screen Download credentials JSON

---

Environment Setup

• Create a .env file:

  EMAIL=your_email@gmail.com
  OAUTH_CREDENTIALS_FILE=Oauth_cred_chatbot_gmail.json
  VERIFY_TOKEN=your_verify_token
  ACCESS_TOKEN=your_whatsapp_access_token
  PHONE_NUMBER_ID=your_phone_number_id
  

  ---

Running the Application

-- Running the App

uvicorn main:app --reload
ngrok http 8000
python gmail.py

Core Components

main.py

• FastAPI webhook server
• Handles WhatsApp events
• Routes messages to LLM

gmail.py

• Fetches unread emails
• Cleans + truncates content
• Sends AI-generated replies

model.py

• Loads Qwen LLM
• Maintains session memory
• Generates responses using chat templates

File	Responsibility
main.py	WhatsApp webhook handler
gmail.py	Email processing & replies
model.py	LLM + session memory

---

Performance Notes

• First run loads model into memory (~5–10 sec)
• CPU inference: ~2–4 sec per response
• Input size significantly affects latency

---

Tech Stack

• FastAPI
• Qwen 0.5B
• Flask
• Gmail API
• WhatsApp API
• ngrok

---