Journal & Graph AI is a high-fidelity personal growth platform that transforms unstructured thoughts and voice recordings into a structured, hierarchical knowledge base. It leverages the Google Gemini API to classify human effort and visualizes progress through a stable Directed Acyclic Graph (DAG).
Project URL: https://self-statistics-system-v1.web.app
The application organizes knowledge into three distinct functional layers, styled as a real-life RPG status system:
- Actions (Emerald #10b981): Granular activities extracted from journal entries (e.g., "Debugging", "Squats").
- Skills (Amber #f59e0b): Intermediate competencies that encapsulate groups of actions (e.g., "Frontend Engineering").
- Characteristics (Indigo #4f46e5): High-level human traits representing fundamental potential (e.g., "Intellect", "Vitality").
The "Brain" is dynamic. Gemini suggests weights for relationships; the system uses a Learning Rate (0.01) to adjust existing edge weights over time, allowing the hierarchy to adapt to your evolving habits.
Effort is quantified as EXP and propagated using Path-Weighted Cumulative Averaging:
- EXP flows upward from Actions to Characteristics.
- Normalization prevents "Domain Inflation," ensuring root attributes reflect average intensity across sub-trees.
A hierarchical feed (Year > Month > Day) for voice and text ingestion. Features detailed "Neural Impact Analysis" for every entry, showing exactly where your effort went. Includes a Canvas-based Oscilloscope for real-time voice feedback.
A stable, interactive DAG visualization designed for semantic clarity and structural persistence. Supports multi-node selection, relationship path highlighting, and Ultra-Rigid position snapping for a deterministic UX.
A character sheet displaying global levels, total EXP, and a "Power Grid" of core attributes. Tracks daily and yesterday's EXP deltas via hierarchical metadata.
A persistent Discord-style interface for managing profile metadata and external connections:
- AI Configuration: Fine-tune classification temperature, model selection, and transcription feedback.
- Data Portability: Full JSON Backup & Restore of the IndexedDB state.
- Webhooks: Real-time JSON broadcasts of processed entries.
- Obsidian: Local REST API integration to sync entries as Markdown notes.
Low-level access to the application engine. Features batch data injection (AI, Manual, Complex, and Brain datasets), an experience injector for math verification, and a raw data browser for IndexedDB inspection.
- Framework: React 19 (Bulletproof Architecture).
- AI Engine: Google Gemini (
gemini-3-flash-preview). - Visuals: D3.js with stable hierarchical layouts and rigid snapping.
- Persistence: Local-First IndexedDB.
- Connectivity: PWA-ready with Service Worker and Manifest support.
- Brain Logic: Topology and propagation utilities live in
lib/soulTopology.
For deep-dives into the implementation, see the /documentation directory.