Built by Aptiva AI for the lablab.ai Hackathon 2025
ReplyMate AI Agent is a real-time conversational intelligence system that enables secure, automated communication with built-in fraud detection and contextual memory. It’s designed for businesses that value smart, responsive, and ethical AI-powered messaging.
- AI-powered Responses using GPT-4o
- Fraud Detection with hybrid rule-based + LLM techniques
- Smart Context Memory via LangChain and Redis
- Multi-Platform Messaging (WhatsApp Business API, Webchat, Email)
- Real-Time Backend with FastAPI / Node.js
- Frontend Dashboard built using React.js
- Scalable Databases using MongoDB and PostgreSQL
- Revisit Consent Button to let users update data-sharing preferences anytime
| Frontend | Backend | AI/NLP | Infra & Database |
|---|---|---|---|
| React.js | FastAPI / Node.js | GPT-4o, LangChain | Redis, MongoDB, PostgreSQL |
| Name | Role | GitHub |
|---|---|---|
| Eesha Tariq | AI Engineer / Team Lead | codewithEshaYoutube |
| Zeeshan Tariq | Data Scientist | zeeshantariqpkn |
| Ramspheld Onyango | Python Developer | RamspheldOnyangoOchieng |
| Najma Razzaq | Data Scientist | najmarazzaq761 |
| Deepak Kumar | Software Design Engineer | Deepak-tekchandani |
| Muhammad Ehtisham | Software Engineer | Ehtisham3397 |
ReplyMate AI Agent is built using a modern, open-source-centric tech stack designed for real-time, secure, and intelligent AI communication systems.
- React.js
A powerful JavaScript library used for building the dashboard and user-facing components. It enables dynamic rendering, state management, and seamless integration with APIs.
✅ Open-source
-
FastAPI
A high-performance Python web framework used for serving AI models, API routing, and backend logic.
✅ Open-source -
Node.js
Used for handling additional backend microservices, particularly integration with messaging APIs like WhatsApp.
✅ Open-source
-
GPT-4o (OpenAI)
The core LLM (Large Language Model) powering ReplyMate's intelligent and contextual responses. While OpenAI is proprietary, the integration logic and how it's called is customizable. -
LangChain
An open-source framework for chaining LLM calls and maintaining smart memory between user sessions. This helps in generating responses that consider prior messages.
✅ Open-source
- Redis
In-memory data store used for caching conversation states, user sessions, and memory embeddings to allow real-time updates and responsiveness.
✅ Open-source
-
MongoDB
NoSQL database for storing chat histories, user profiles, feedback, and logs.
✅ Open-source -
PostgreSQL
Used for structured data such as configuration settings, audit logs, and team permissions.
✅ Open-source
ReplyMate AI includes a hybrid fraud detection system combining:
-
Rule-Based Logic:
Custom-defined rules such as keyword filtering, blacklist checking, and frequency analysis. For example, repetitive requests for personal data or financial details can trigger warnings. -
LLM-Based Analysis (using GPT-4o):
Suspicious messages are also passed through an LLM prompt with zero-shot classification prompts to evaluate intent and context. This provides semantic-level fraud detection. -
Scoring System:
Every message receives a fraud score based on rule matches and LLM evaluation. Messages exceeding a threshold are flagged or auto-replied with a warning. -
Consent-Aware Blocking:
Users can revisit their consent through the "Revisit Consent" feature to manage data-sharing and report abuse proactively.
-
🚀 Live Web App:
ReplyMate AI - Streamlit Deployment -
✅ Lablab.ai Submission:
ReplyMate AI — AI Genesis Hackathon by Aptiva AI -
🧑💻 GitHub Repository:
https://github.com/codewithEshaYoutube/ReplyMate-Agent
# Clone the repository
git clone https://github.com/YourRepo/ReplyMate-AI-Agent.git
cd ReplyMate-AI-Agent
# Backend setup
cd backend
pip install -r requirements.txt
uvicorn main:app --reload
# Frontend setup
cd ../frontend
npm install
npm start