🤖 Wesley Bot — AI-Powered WhatsApp Assistant & RAG

🚀 Overview

Wesley Bot is an advanced virtual assistant for WhatsApp, developed to provide intelligent, contextualized, and multimodal responses. It uses an architecture based on Python and FastAPI, integrating with the WhatsApp ecosystem via Evolution API and powered by Google Gemini.

The project's highlight is the implementation of RAG (Retrieval-Augmented Generation), allowing it to "learn" and query specific documents (PDFs, TXT) in real-time for precise answers.

🎯 Value Proposition

• Contextual Memory: Utilization of FAISS and Embeddings for semantic search in documents (RAG).
• Multimodal AI: Processes and generates both text and audio (via gTTS).
• Agentic Flow: Automatically detects user needs (responds via text, sends audio, or exports data to a spreadsheet).
• Robust Integration: Stable communication via Evolution API.
• Production-Ready: Fully containerized with Docker.
• Deploy: Running on Oracle Cloud Always Free (Always Free Tier instances), on a dedicated VPS for continuous bot availability.

🏗️ System Architecture

%%{title: "Wesley Bot WhatsApp Assistant Architecture"}%%
graph TB
    subgraph "Client Layer"
        A[WhatsApp User]
    end

    subgraph "Integration Layer"
        B[Evolution API]
    end

    subgraph "Core Backend (FastAPI)"
        C[FastAPI Backend]
        F[gTTS - Text-to-Speech]
        G[Openpyxl - Excel Export]
    end

    subgraph "AI & Data Layer"
        D[Google Gemini - AI]
        E[(FAISS - Vector DB)]
        H[RAG Documents]
    end

    A <--> B
    B <--> C
    C --> D
    C --> E
    C --> F
    C --> G
    E --> H

🛠️ Tech Stack

• Python 3.11 - Base language for its versatility in AI and automation.
• FastAPI - Modern, high-performance web framework for the bot's API.
• Evolution API - High-performance connector for the WhatsApp protocol.
• Google Gemini - Next-generation language model for agentic processing.
• FAISS - Facebook library for efficient semantic vector search.
• gTTS (Google Text-to-Speech) - Converts AI responses to audio for the user.
• Docker - Containerization to ensure consistency across environments.

🎯 Features

1. RAG-Enabled Assistant

• The bot can ingest documents and base its answers on their content.
• Semantic search: finds the correct information even if the user doesn't use the exact words from the document.

2. Automated Spreadsheet Flow

• Capable of collecting information during the conversation and automatically exporting it to Excel (XLSX) files.
• Ideal for lead management, order registration, or checklists.

3. Voice Responses

• When requested or configured, the bot converts its text responses into natural-sounding audio, sending the voice message directly on WhatsApp.

4. History Management

• Maintains conversation context, allowing for fluid dialogues and references to previous messages.

🚀 How to Run

Prerequisites

• Docker and Docker Compose
• Google Gemini API Key
• Running instance of Evolution API

Installation

1. Clone the repository.
2. Configure the .env file with your credentials (Gemini Key, Evolution API URL/Token).
3. Spin up the environment with Docker:

   docker-compose up -d --build
   

Deploy (Production)

The bot is in production on Oracle Cloud Always Free (Always Free Tier), using VPS instances to run the containers (FastAPI backend + Evolution API when applicable). The Always Free tier allows keeping the assistant available 24/7 with no infrastructure cost.

Empowering WhatsApp communication with Artificial Intelligence.