RedFlagger is an emotionally intelligent AI-powered analyzer for WhatsApp chats. It goes beyond generic sentiment analysis by identifying manipulation patterns, sudden emotional spikes, and potential red flags in day-to-day conversations.
Built with LangChain, OpenAI- GPT-4o, and Streamlit, it transforms raw exported .txt WhatsApp files into insightful, color-coded visualizations that surface the emotional truth beneath the surface of your chats.
- ✅ WhatsApp Chat Loader & Cleaner (media-aware)
- 🧠 LLM-powered Sentiment Analyzer (with fine-grained classification)
- 📉 Sentiment Timeline Chart (track emotional fluctuations)
- 🌩️ Spike Detection & Red Flag Alerts
- ☁️ Color-coded Message Cloud
- 🧪 Ready for Agentic workflows and RAG-based enhancements
- 🧾 Input Module: User uploads .txt export of WhatsApp chat. Parser splits messages: sender, timestamp, content.
- 🧮 Preprocessing & Chunking: Clean metadata, remove system messages. Chunk into: Per day / week Per topic turn (based on shifts in conversation)
- 🧰 Analysis Engines (Chains) 🔴 Red Flag Detector
- Gaslighting
- Breadcrumbing
- Guilt-tripping
- Blame-shifting
- Love bombing
- Silent treatment
- Output: Flag + explanation + message reference.
- Simple Streamlit UI: Upload chat Run analysis Visual output: flags, charts, downloadable report
PythonLangChainOpenAI GPT (gpt-4o-turbo)(or configurable)StreamlitMatplotlib / SeabornPandas
- Export your WhatsApp chat as
.txtfile (without media). - Launch the Streamlit app:
streamlit run main.py
- Upload chat file, view timeline, red flags, and insights.
⚠️ Media placeholders (like"<Media omitted>") are intelligently handled.
- Relationship Health Score
- Toxicity Detection
- Persona-Based Behavior Summaries
- Support for multiple chat formats
MIT License. Open for contributions!