AI-powered emotional diary that understands and comforts you
- 🧠 AI Emotion Analysis — Real-time emotion detection powered by Groq (gpt-oss-120b)
- 💬 Empathetic AI Characters — Personalized comfort messages from AI companions
- 🌱 Emotion Calendar — Visualize your emotional journey as growing plants
- 📊 Statistics & Trends — Emotion charts, keyword analysis, and weekly insights
- 📸 Image Analysis — Detect emotions in photos via Vision API
- 🔔 Smart Notifications — Morning encouragement & evening care across 2 channels
- 🔒 Privacy-First — No account, no cloud sync. Entries live on your device and go out only for the AI analysis itself.
- ✅ 1,814 tests passing — the test suite is roughly the size of the production code
- 📝 Diary auto-save — in-progress entries and attached photos now survive backgrounding, back-press, and process death; restored via a banner, kept for 7 days
- ☎️ Crisis hotline unified to 109 — Korea merged its suicide-prevention line in Jan 2024; all 6 call sites now match
- 🔐 Removed unauthenticated admin endpoints — 3 public Cloud Functions HTTP handlers deleted (they could push to every topic subscriber)
- 🖼️ Groq Vision rate-limit handling — one downscaled image per request, with text-only fallback when the token budget is exceeded
Previous Updates (v1.4.60 – v1.4.63)
- 🔒 Privacy policy rewritten — replaced a generic template with the app's actual data flow (Groq, Firebase, on-device storage)
- 🐛 FCM
{name}placeholder leak fixed — traced past the client to the server-side message templates, where the real source was - 🧪 Non-vacuous regression tests — new tests are now verified to fail against the pre-fix code before being accepted
- ♻️ Health Check refactor — 57 files restructured with no user-facing change
- 📏 Repository-wide format cleanup — 53 files realigned to the Dart 3.7+ tall-style formatter
Your mind belongs to you.
| Item | Policy |
|---|---|
| Storage | Local SQLite on your device — no account, no cloud sync |
| AI Analysis | Your entry text, plus one attached photo, is sent to the Groq API (US) to be analyzed. No name, email, or account is attached. |
| Analytics | Firebase receives the emotion score, energy level, and the first 50 characters of the action suggestion — never the entry itself. |
| Deletion | Instant full deletion from Settings |
See Privacy Policy for details.
| Category | Technology | Version |
|---|---|---|
| Framework | Flutter / Dart | 3.38.x / ^3.10.1 |
| State | Riverpod | 2.6.1 |
| Database | SQLite (sqflite) | 2.3.3 |
| Firebase | Analytics, Crashlytics, FCM | 3.8.0+ |
| Routing | go_router | 17.0.1 |
| AI (text) | Groq API | openai/gpt-oss-120b |
| AI (vision) | Groq API | qwen/qwen3.6-27b |
| Charts | fl_chart | 0.68.0 |
┌──────────────────────────────────────────┐
│ Presentation │
│ Providers (Riverpod) + Widgets │
├────────────────────┬─────────────────────┤
│ ▼ │
│ Domain │
│ Entities, UseCases, Repo Interfaces │
├────────────────────┬─────────────────────┤
│ ▲ │
│ Data │
│ Repo Impl, DataSources, DTOs │
└──────────────────────────────────────────┘
Layer rules: presentation → domain ← data
(domain has zero external dependencies)
- Flutter 3.38.x / Dart 3.10.x
- A Groq API key
# Clone
git clone https://github.com/kaywalker91/MindLog.git
cd MindLog
# Install dependencies
flutter pub get
# Generate code (freezed, json_serializable, etc.)
dart run build_runner build --delete-conflicting-outputs
# Run
flutter run --dart-define=GROQ_API_KEY=your_key# Release App Bundle
flutter build appbundle --release --dart-define=GROQ_API_KEY=your_key
# Release APK
flutter build apk --release --dart-define=GROQ_API_KEY=your_keylib/
├── core/ # Config, services, theme, constants, utilities
├── data/ # Repository implementations, DataSources, DTOs
├── domain/ # Pure Dart: entities, repository interfaces, use cases
├── presentation/ # Providers, Screens, Widgets
└── main.dart
This project is built with AI coding assistants, and the working agreement for that is
version-controlled rather than improvised per session. .claude/ holds the
architecture and testing rules the assistant must follow, the automation commands used for
recurring work, and the engineering notes written up after each non-obvious bug — including
why a regression test that passes can still be worthless.
# Run all tests with coverage
./scripts/run.sh test
# Full quality gates (lint + format + test)
./scripts/run.sh qualityCoverage targets: unit ≥ 80%, widget ≥ 70%
Bug reports, feature requests, and pull requests are welcome!
See CONTRIBUTING.md for guidelines.
See CHANGELOG.md for full release notes.
MIT License — Copyright (c) 2024 kaywalker91