2.9 KiB
2.9 KiB
Master User Story: Distributed, Local-First Productivity Engine
Canonical Reference: docs/ARCHITECTURE/04-master-user-story.md
This document provides workflow-level summaries of the three primary execution domains. For full acceptance criteria and technical specifications, refer to the canonical master user story document above.
1. Multi-Domain Ingestion Workflows
The system automatically captures inputs from multiple distinct channels without requiring manual task entry:
- IMAP Email Ingestion: Background daemon polls personal, engineering, and commercial IMAP accounts over TLS. Flagged emails or messages forwarded to internal parse addresses are extracted as Markdown notes with linked tasks.
- Global Desktop Jots: System-wide hotkeys (Windows/Linux) invoke a lightweight command palette for instant plain-text jots. Android home-screen widgets and Share Intent receivers forward text, URLs, and voice notes.
- Local File Watchers: Background filesystem watchers monitor dedicated folders. Markdown files with
- [ ]checkboxes automatically parse into tasks. PDFs and design briefs parse into linked project notes. - AI Processing Pipeline: All raw inputs pass through the local NLP engine (regex + optional Ollama LLM) for namespace routing, task/note splitting, date extraction, and Eisenhower priority scoring.
2. Serverless P2P CRDT Synchronization
Data state changes are tracked via Hybrid Logical Clocks (HLC) and append-only CRDT deltas:
- Local-First: All data is written to the local SQLCipher database first. The app is fully functional offline.
- Peer Discovery: Nodes discover each other on LAN via mDNS (zero-config) and across WAN via libp2p Kademlia DHT with NAT hole-punching.
- Delta Sync: On reconnection, peers exchange vector clocks, compute missing deltas, compress and encrypt them, and stream over QUIC/Noise XK channels.
- Convergence: Automerge CRDT deterministic merge guarantees eventual consistency across all personal nodes (Windows, Android, Headless Server) without central server coordination.
3. Multi-View Display Execution
The Flutter presentation layer renders local database state into four specialized execution environments:
- Eisenhower Priority Matrix: Four-quadrant board auto-populated by computed urgency-importance scores. Tasks automatically migrate between quadrants as deadlines approach.
- Calendar Time-Blocking Grid: Drag-and-drop scheduling grid with CalDAV overlay for external calendar integration. Tasks reserve focus blocks with real-time duration adjustment.
- Kanban / Gantt Views: Status-column Kanban board with swimlane grouping; interactive Gantt timeline with critical path highlighting and fractional sub-task nesting.
- Notes & Knowledge Workspace: Rich Markdown editor with bidirectional task linking, project tree navigation, and Daily Jots aggregation.