~/orlix/CHANGELOG
$ git log --oneline --graph

Changelog

Release history for Orlix — personal AI operating system.

v0.5
building now
● in progress

Evolve — policy learning from outcomes

The loop closes. Orlix learns from what actually happened — not just what you told it.

  • Automatic policy evolution from verified outcomes
  • Pattern detection across receipt history
  • Anomaly flagging when behavior diverges from expected
  • Policy diff view before any automatic change is applied
  • Adaptive notification frequency based on signal importance
v0.4
2026-04-10
✓ shipped

Verify — outcome tracking

Actions that execute are not the same as goals that succeed. v0.4 closes the loop.

  • Outcome verification engine — checks actual results against stated intent
  • Receipt system: every action logs intent, policy, result, and rollback state
  • 30-day rollback window for all supervised actions
  • Goal drift detection — alerts when a goal hasn't progressed in configurable window
  • Discrepancy queue — flagged actions held for human review
  • JSON export format for all receipts (readable and portable)
v0.3
2026-02-14
✓ shipped

Act — supervised execution

Orlix takes its first real actions — within explicit approval workflows.

  • Five authority tiers: observe, suggest, confirm, supervised, autonomous
  • Per-domain tier overrides (email vs. calendar vs. files)
  • Approval workflow with inline context — one-click confirm or deny
  • Linear, Notion, and GitHub Issues adapters
  • Local filesystem adapter (read/write within configured paths)
  • SMTP adapter for supervised email drafting and sending
v0.2
2025-11-20
✓ shipped

Decide — rules-based decision engine

The world model gains a conscience — a policy engine that evaluates what should happen before anything does.

  • Policy engine with priority weighting and authority checks
  • Versioned policy system — every change tracked, any version restorable
  • Goal registry with deadlines, progress tracking, and blockers
  • Conflict resolution when multiple policies apply to the same signal
  • Policy diff UI — review before saving any change
  • CLI: orlix policy --list, --add, --rollback
v0.1
2025-08-05
✓ shipped

Observe — persistent world model

The foundation. A local-first memory system that persists across sessions and stays with you regardless of which AI you use.

  • Local-first fact graph with source tracking and confidence scores
  • Cross-session persistence — memory survives model switches and restarts
  • Google Calendar and Apple Calendar adapters
  • IMAP adapter for reading (not acting on) email signals
  • Notes.app and Markdown file ingestion
  • JSON export: full world model in portable, human-readable format
  • Encrypted local credential store
  • CLI: orlix init, orlix status, orlix memory --export
v1.0
late 2026
○ planned

Full symbiotic loop

All five phases running continuously and self-improving. The system becomes a true personal operating system.

  • Multi-device sync with end-to-end encryption
  • Team / shared-context mode for small organizations
  • Plugin SDK for custom adapters
  • Mobile companion app (iOS / Android)
  • Web-based receipt and policy dashboard