Multi-Agent WhatsApp Conversational & Workflow Engine
May 2026 - Current (4 months)
This project is an Asynchronous, AI-Powered WhatsApp Conversational Workflow Engine designed to orchestrate complex user interactions (such as stateful multi-step questions, lists, timers, and conditional flows) over WhatsApp.
Operating as an event-driven webhook handler built for stateless, horizontally scalable compute environments, the engine seamlessly integrates generative AI (Google Gemini, OpenAI, and Groq) directly into its execution logic:
Multi-Modal AI Validation: It leverages Gemini 2.5 Flash's vision capabilities to act as a validation gate for payment receipts (images/PDFs), performing real-time OCR, validating bank names and expected transaction values, and managing retry thresholds.
Semantic Search & RAG: It uses LLMs for