ROADMAP.adoc

dictask Roadmap

Table of Contents

• Current Status
• Milestones
  • v0.1.0 — Prototype (MVP)
  • v0.2.0 — Automated Pipeline
  • v0.3.0 — Intelligence
  • v0.4.0 — Deployment & Hardening
  • v1.0.0 — Stable Release
• Future Directions

Current Status

Design phase — architecture specified, implementation starting.

Milestones

v0.1.0 — Prototype (MVP)

• ❏ SQLite schema finalised with Idris2 ABI proofs

• ❏ Haskell megaparsec parser: basic intent extraction (task title, deadline, priority keywords)

• ❏ Rust ingest: udev detection + local archive with SHA-256 checksums

• ❏ Manual pipeline: ingest → transcribe → parse → store (no automation)

• ❏ Basic view generation (Markdown task list from SQLite)

v0.2.0 — Automated Pipeline

• ❏ systemd service for automatic ingest on recorder insertion

• ❏ Whisper/Vosk integration for offline transcription

• ❏ Confidence scoring with configurable thresholds

• ❏ Review queue for low-confidence items

• ❏ Cloud backup with retry/quarantine

v0.3.0 — Intelligence

• ❏ Semantic deduplication (fuzzy matching + context)

• ❏ Priority scoring formula: urgency * 0.5 + importance * 0.3 + deadline_proximity * 0.2

• ❏ Tentative deadline resolution ("after Easter" → flag for review)

• ❏ Contradictory update detection

v0.4.0 — Deployment & Hardening

• ❏ Ansible playbook for full local setup

• ❏ Terraform for cloud provisioning

• ❏ Full audit trail (audio hash → transcript → parse → task)

• ❏ Privacy hardening (redaction, encrypted storage)

• ❏ Containerised deployment via Stapeln

v1.0.0 — Stable Release

• ❏ All pipeline stages idempotent and replayable

• ❏ Comprehensive test suite (HSpec, cargo test, property-based)

• ❏ PanLL panel for review queue (if applicable)

• ❏ Documentation complete

• ❏ Success metrics dashboard

Future Directions

• Train a model on historical edits to refine confidence scoring

• VeriSimDB backing store for federation/replication

• Multi-format export (Org-mode, Taskwarrior, iCal)