Skip to content
This repository was archived by the owner on Jul 18, 2026. It is now read-only.

AVE 1.0.5

Choose a tag to compare

@lemberalla lemberalla released this 24 Jun 18:13

AVE 1.0.5

This is a large patch release focused on AVE's editing brain: better dirty-brief planning, stronger Ask/Plan routing, and deeper AI control over native graphics and timeline tools.

Highlights

Agentic edit planning

  • Added a structured read/evaluate/refine loop for edit planning.
  • Improved support for long messy briefs, campaign-style requests, and multi-sequence deliverables.
  • Planner can now recover executable timeline actions from richer creative plans instead of stopping at narrative-only output.
  • Expanded validation around edit-plan execution so approved plans are more likely to become real timeline changes.

AI graphic styling

  • Added AI-accessible controls for text and graphic styling.
  • AVE can now reason about and apply typography, color, fills, shadows, strokes, placement, and layer targeting through native tools.
  • Improved support for title cards, lower thirds, CTA cards, end cards, and existing graphic layer edits.
  • Added safer targeting for graphic layers so styling changes land on the intended layer.

Ask mode routing

  • Ask mode now uses structured route decisions instead of guessing from assistant wording.
  • Plan preview cards are only shown when the runtime explicitly decides a plan handoff is needed.
  • Reduced misleading "Switch to Plan mode" prompts during ordinary clarification.
  • Improved Ask-to-Plan handoff so Plan mode receives the actual user intent instead of a stale or incomplete prompt.

Assistant provider cleanup

  • Removed unsupported/unstable Gemini CLI and Antigravity CLI providers from AVE.
  • Kept the supported assistant paths focused on Claude CLI, Codex CLI, Qwen CLI, local models, and BYO API.
  • Updated onboarding, settings, and provider catalog behavior to match the supported provider list.

Reliability

  • Added the edit read loop implementation and tests.
  • Expanded planner, local AI tool, graphic styling, and tool-executor test coverage.
  • Improved plan-action validation and local tool routing around agentic edits.
  • Bumped build number to 6.

Verification

  • Built production renderer and render worker.
  • Built signed macOS arm64 DMG and zip artifacts.
  • Notarized and stapled the app/DMG with Apple notary service.
  • Verified package contents to exclude marketing demo output, docs, scripts, release output, env files, private keys, and secure local stores.
  • Verified Gatekeeper acceptance for the DMG and app bundle.