Skip to content

AION v1.5.0 — Capability Awareness, Self-Improvement & Multi-Channel UI

Latest

Choose a tag to compare

@xynstr xynstr released this 03 Apr 02:57
· 2 commits to main since this release

What's new in v1.5.0

This is the largest release since v1.0.0 — two parallel development lines merged.

Architecture

  • aion.py modularized — split into core/ package: aion_config, aion_character, aion_permissions, aion_prompt, aion_providers — each independently hot-reloadable
  • core/aion_progress.py — per-task progress reporting for real-time frontend progress bars
  • All tools always visibletier_threshold=2 default; desktop, browser, audio tools always in API schema — eliminates hallucinated tool names and wasted retry turns
  • Grouped Capability Index — system prompt contains a semantic, auto-generated tool index grouped by prefix (DESKTOP, AUDIO, BROWSER…)

Self-Improvement System

  • record_mistake tool — AION records mistakes to mistakes.md; last 5 entries injected into every session
  • lookup_rule tool — keyword/section search over prompts/rules.md
  • Boot maintenance sessionboot_session plugin spawns a silent background session on startup after ≥1h offline; reviews mistakes, character, todos
  • Offline duration hint — system prompt shows how long AION was offline
  • Doc freshness warning — warns when core .py files are newer than AION_SELF.md
  • End-of-conversation protocol — mandatory wrap-up rules: update_character, reflect, record_mistake, CHANGELOG check
  • max_rules_chars: 35000 — full 27 KB rules.md now always loaded (was truncated at 12 KB)

New Tools & Plugins

  • desktop_set_window_state — minimize, maximize, restore or close any window by partial title match
  • boot_status — last boot time, offline duration, maintenance status
  • record_mistake / lookup_rule — persistent learning & rule lookup
  • boot_session plugin — startup maintenance session
  • docx_to_speech plugin — convert Word documents to MP3/WAV via edge-tts or sapi5

Multi-Channel UI

  • Channel switcher — pill tabs for web/Telegram/Discord/Slack with unread counts
  • Detach-task (⊞) — offload long tasks to background while keeping chat responsive
  • Mood system — visual mood badges and bubble tinting
  • Activity log — tool execution timeline
  • Update banner — notification when a new version is available
  • Real-time progress bars — for long-running tool operations

Plugin Updates

  • anthropic_provider — Claude 4.x models (opus-4-6, sonnet-4-6, haiku-4-5)
  • gemini_provider — Function Calling, thinking budgets, vision — all stable
  • audio_pipeline — OGG/MP3/WAV/FLAC/WebM via Faster Whisper; multi-engine TTS
  • telegram_bot — feature parity: isolated history, approval buttons, voice, photos

Documentation & Maintenance

  • MAINTENANCE.md — change-propagation checklists for every change type
  • docs/messaging.md — full integration guide (Telegram, Discord, Slack, audio, MCP)
  • hub-repo-template/ — distributable plugin repo framework with GitHub Actions

Fixes & Cleanup

  • Wakeup delivery via config.json poll (SSE queue was unreliable)
  • confirmed not required in desktop_set_window_state
  • Optional[float] in boot_session.py for Python 3.9 compatibility
  • thinking_budget=512 for Gemini 2.5 wakeup (0 was invalid)
  • Removed clio_reflection stub plugin and dead integration code
  • Removed duplicate runtime files from core/ package