AI Message Notification Router

An AI-powered notification routing system that intelligently classifies incoming WhatsApp messages into Notify, Digest, or Mute using a modular multi-agent architecture and context-aware reasoning.

---

Project Overview

This project was developed as a solution for the HackerRank Orchestrate – Message Notification Router challenge.

The system analyzes incoming WhatsApp messages using message content, user behaviour, historical interactions, business metadata, and contextual information to determine the most appropriate notification strategy.

The objective is to minimize unnecessary interruptions while ensuring important messages receive immediate attention.

---

Features

• Multi-Agent Architecture
• Context-Aware Decision Making
• Personalized Notification Routing
• Rule-Based Reasoning Engine
• Knowledge Base for Historical Context
• Policy Validation Layer
• Evaluation Pipeline
• CSV-based Prediction Generation

---

Architecture

Incoming Messages
        │
        ▼
Dataset Loader
        │
        ▼
Knowledge Base
        │
        ▼
Message Intake Agent
        │
        ▼
Context Retrieval Agent
        │
        ▼
Rule-Based Reasoning Agent
        │
        ▼
Policy Validator
        │
        ▼
Output Formatter
        │
        ▼
output.csv

---

Project Structure

code/
│
├── agents/
│   ├── message_intake.py
│   ├── context_retrieval.py
│   ├── rule_based_reasoning.py
│   ├── policy_validator.py
│   └── output_formatter.py
│
├── knowledge/
│   ├── loader.py
│   └── knowledge_base.py
│
├── models/
│   ├── structured_message.py
│   ├── context_bundle.py
│   ├── reasoning_state.py
│   └── decision_result.py
│
├── services/
│   ├── prompt_builder.py
│   ├── groq_service.py
│   └── json_parser.py
│
├── evaluation/
│   └── main.py
│
├── main.py
└── requirements.txt

---

Decision Workflow

For every incoming message the system:

1. Loads structured message data.
2. Retrieves user, business, group and historical context.
3. Applies rule-based reasoning using contextual evidence.
4. Validates the decision against notification policies.
5. Generates the final routing decision.
6. Writes predictions to output.csv.

---

Notification Actions

Action	Description
Notify	Important messages requiring immediate attention
Digest	Useful but non-urgent notifications
Mute	Low-value, repetitive, spam or suspicious messages

---

Technologies Used

• Python
• Pandas
• Pydantic
• Rule-Based AI
• Multi-Agent Architecture

---

Installation

Clone the repository:

git clone https://github.com/Harikarthik7124/Message-Notification-Router.git

Install dependencies:

pip install -r requirements.txt

---

Running the Project

Place the provided dataset in the project directory and run:

python main.py

The generated predictions will be saved as:

output.csv

---

Evaluation

Run the evaluation script:

python evaluation/main.py

The evaluation reports:

• Number of predictions
• Missing records
• Duplicate IDs
• Decision distribution
• Average confidence

---

Future Improvements

• Integrate LLM-based reasoning
• Image OCR support
• Voice note transcription
• Hybrid Rule + LLM decision engine
• Confidence calibration using model feedback

---

About

This repository contains my implementation of the Message Notification Router challenge from the HackerRank Orchestrate hackathon. The architecture, implementation, and decision pipeline represent my own solution developed for the challenge.

---

License

This repository is intended for educational and portfolio purposes.