Voice Journal

🔍 The Problem

Journaling is powerful, but typing on a phone kills the flow. Thoughts come fastest when you're walking, commuting, or lying in bed—moments when a keyboard is the last thing you want.

Voice memos solve the capture problem but create a new one: recordings pile up, unsearchable and unreviewed. You end up with hundreds of audio files and zero actionable insights.

💡 The Solution

A mobile-first voice journal that turns speech into structured, searchable entries through an AI pipeline: Record → Transcribe → Summarize → Tag → Analyze.

Every entry gets an AI-generated summary, mood tag, and topic classification—making weeks of thoughts instantly searchable and reviewable through weekly insight reports.

🛠️ Technical Implementation

🎙️ Recording Engine

• Built on Expo AV for cross-platform audio capture
• Pause/resume support for natural, unstructured thoughts
• Background recording with lock-screen controls
• Configurable audio quality and format settings

🧠 AI Processing Pipeline

• Step 1 — Transcription: OpenAI Whisper API converts audio to text with high accuracy
• Step 2 — Summarization: GPT-4o-mini condenses transcripts into concise journal entries
• Step 3 — Auto-Tagging: AI extracts topics, people, and themes from content
• Step 4 — Mood Detection: Sentiment analysis classifies emotional tone per entry

💾 SQLite Schema & Migrations

• Normalized schema: entries, tags, entry_tags, weekly_summaries
• Version-tracked migrations for safe schema evolution
• Full-text search index on transcripts and summaries
• Efficient pagination for large journal histories

🔐 Privacy & Security

• Biometric authentication (Face ID / fingerprint) on app launch
• All data stored locally on-device via SQLite
• Audio files encrypted at rest
• Zero cloud dependency—no account required

🎯 Key Features

• 🎙️ Voice Recording: Tap-to-record with pause/resume and waveform visualization
• 📝 AI Transcription: Whisper-powered speech-to-text with punctuation and formatting
• 📋 Smart Summaries: GPT-4o-mini distills rambling thoughts into clear entries
• 🏷️ Auto-Tagging: Topics, people, and themes extracted automatically
• 😊 Mood Detection: Emotional tone tracked across entries for pattern recognition
• 🔍 Full-Text Search: Find any thought by keyword across all entries
• 📊 Weekly Insights: AI-generated analytics on mood trends, top topics, and journaling streaks
• 📴 Offline-First: Works without internet; AI processing queues until connected

📊 Technical Highlights

Metric	Detail
🧠 AI Pipeline Steps	4 (Transcribe → Summarize → Tag → Mood)
💾 Database Tables	4 (entries, tags, entry_tags, weekly_summaries)
🔍 Search Latency	< 100ms full-text across all entries
🔐 Auth Method	Biometric (Face ID / Fingerprint)
📴 Offline Support	Full — queues AI processing for sync

📱 Cross-Platform: iOS & Android

🎯 Why This Matters

This project demonstrates:

• 📱 Mobile Development: Production React Native with Expo, handling audio, background tasks, and native APIs
• 🤖 AI Integration: Multi-step pipeline chaining Whisper and GPT-4o-mini for structured output
• 💾 Local-First Architecture: SQLite with migrations, FTS indexing, and encrypted storage
• 🎨 Custom Design System: Purpose-built UI components for a journaling experience
• 🔒 Privacy Engineering: Zero-cloud architecture with biometric auth and on-device encryption

🔧 Stack

• React Native / Expo: Cross-platform mobile framework
• TypeScript: Type-safe development across the entire codebase
• SQLite: Local-first storage with full-text search
• OpenAI Whisper: Speech-to-text transcription API
• GPT-4o-mini: Summarization, tagging, and mood analysis
• Expo AV: Audio recording and playback engine