MindJournal

After stressful events (a failed presentation, rejection, a missed deadline), people journal but receive no structured feedback. There is no lightweight system that detects distorted thinking (catastrophizing, overgeneralization), extracts core negative beliefs, suggests concrete improvement actions, and aggregates recurring cognitive patterns — leading to repeated negative thought loops without measurable self-awareness.

An LLM-powered CBT-lite reflection system that analyzes a journal entry, extracts the core negative belief, cognitive distortion type, and emotion intensity, then generates 3 reframes, 1 guided reflection question, and 2–3 actionable self-improvement steps. Structured outputs are stored in a Cognitive Pattern Module, and a dashboard shows distortion frequency, mood trends, and reframe effectiveness.

• Write a journal entry after a stressful event
• The system extracts core belief, distortion type, and emotion intensity
• Receive 3 reframes, 1 reflection question, and 2–3 action steps
• Structured outputs feed the Cognitive Pattern Module
• Dashboard surfaces distortion frequency, mood trends, and reframe effectiveness over time

• Structured information extraction — text → JSON schema
• Cognitive distortion classification
• Controlled reframe generation
• Context-aware action suggestion

• Frontend & dashboard: Streamlit
• Backend: Python
• Storage: SQLite (local pattern tracking)
• LLM: Gemini API (analysis + chart insight generation)
• Vibe coding: ChatGPT / Gemini