🛡️ ChatCompanion

Problem

Online grooming, cyberbullying, and manipulation in chat conversations are growing threats, especially for young people and vulnerable individuals. Statistics show alarming trends:

• 89% increase in online grooming crimes over six years (UK, 2024)
• 1 in 5 teens (20%) reported experiencing sextortion (2025)
• 55% of teens have experienced cyberbullying in their lifetime
• 55.8% of bullied students did NOT tell a trusted adult
• 20% of minors who experienced online sexual interactions told no one

Existing solutions often require uploading conversations to third-party servers, creating privacy concerns. Many victims don't recognize manipulation patterns until it's too late, and they often don't know where to turn for help.

Motivation

I built ChatCompanion because I believe everyone deserves to understand what's happening in their conversations—especially when someone is trying to manipulate, pressure, or harm them. The tool addresses a critical gap: providing immediate, privacy-first analysis without requiring users to trust third-party services with their sensitive conversations.

This project combines my technical skills with a deep commitment to ethical technology and user privacy. Every conversation analyzed stays completely local—no data leaves the user's device, no third parties are notified, and no tracking occurs. This privacy-first approach is essential when dealing with such sensitive and potentially dangerous situations.

Solution

ChatCompanion is an offline, privacy-first companion that:

• Analyzes locally: All pattern detection happens on the user's device using rule-based detection and optional ML models
• Explains clearly: Uses simple, non-technical language to explain detected patterns (manipulation, bullying, pressure, grooming)
• Assesses risk: Traffic light system (GREEN/YELLOW/RED) provides clear visual feedback on conversation risk levels
• Supports boundaries: Provides concrete phrases and strategies for setting healthy boundaries
• Encourages help: Offers resources and guidance for reaching out to trusted adults or professional support
• Protects privacy: 100% offline operation ensures conversations never leave the user's device

My Responsibilities

Solo Project: ChatCompanion was developed entirely by me as an open-source project focused on privacy, ethics, and helping vulnerable individuals.

• Full-Stack Development: Complete Streamlit application architecture, pattern detection engine, and user interface design
• Pattern Detection: Rule-based analysis system for identifying manipulation, bullying, pressure, and grooming patterns in conversations
• ML Integration: Optional machine learning models for enhanced detection capabilities (~80MB download, completely optional)
• Privacy Architecture: 100% offline-first design ensuring no data leaves the user's device, no third-party tracking, and complete privacy protection
• Documentation & Ethics: Comprehensive documentation, ethics statements, security policies, and clear explanations of tool capabilities and limitations
• User Experience: Traffic light risk assessment system, simple language explanations, boundary-setting guidance, and help resource integration

Impact

ChatCompanion provides a free, open-source tool that helps individuals recognize dangerous patterns in their conversations before situations escalate. By operating completely offline and explaining risks in accessible language, it empowers users to make informed decisions and seek help when needed—without compromising their privacy.