Skip to content

v1.1.2 — Fake-tool audit: 4 made real, 2 deleted

Choose a tag to compare

@github-actions github-actions released this 08 Jul 14:37
· 5 commits to main since this release

Tester-reported fake tools, hunted down for real

Testers found tools that still pretended to work. This release's audit didn't rely on pattern-matching: it executed every tool against a fresh mock Set and diffed a deep state fingerprint before/after — any action tool returning success without actually changing anything (and without an honest advisory flag) was read and judged by hand.

Made real (4)

  • midigate__generate_gate_pattern — used to return a Math.random() note count without writing a single note. Now writes an actual MIDI gate-pattern clip: deterministic euclidean spread by density, swing on offbeats, accents on downbeats. Undoable via Edit History.
  • midigate__set_gate_pattern — used to create an empty junk track. Now writes your binary pattern (1/x on, 0/./- off) as real notes in a real clip. The Trance Gate panel now genuinely writes what you draw.
  • organizer__group_scenes — was analysis dressed as an action. Now classifies scenes AND (opt-in apply_names) renames them in place with a [Section] prefix — real scene.name writes, idempotent.
  • clips__auto_tag_clips — claimed autoTagged:true while writing nothing. Now honest analyze-by-default, with opt-in write_names that appends #tags to each clip's real name (Live has no separate tag field — the description says so).

Deleted (2) — nothing real was possible behind the claim

  • midigate__gate_to_audio — created an empty audio track claiming rendered:true. Real audio gating needs the in-Live render pipeline; use the pattern clip to sidechain a Gate device instead.
  • temposync__sync_tracks_to_tempo — set the global tempo (duplicating set_tempo) then fabricated per-track "synced" results. The SDK has no per-track sync operation, so the tool is gone.

Tool count is now an honest 380 (was 382), refreshed across the app, docs, landing page and catalog.

Verification

299/299 smoke tests (7 new regression checks: real note write-back verified by an independent module, no-RNG determinism, deleted tools absent, real renames), run 3×. The dynamic audit was re-run after the fixes and comes back clean — every remaining flag was individually reviewed as a justified false positive. Browser-verified end to end: the Trance Gate panel applied a pattern and notation__get_clip_notes read the 8 real notes back from the clip — before this fix it would have read 0.