Flip publishing from audioif to audiocomponents

[bdbarnett]

Deferred deliberately at seed time (2026-09-01). This repository publishes
nothing; audioif still owns pydevices-audioinstruments,
pydevices-audioeffects and the two MIP index entries, and keeps its own
copies of both packages.

Recording the checklist here rather than in a README so it does not go quiet.
None of this should happen until Brad says so — most likely after the
accuracy rewrite, not before.

Human steps (Brad's alone)

• Install the pydevices-release-automation GitHub App on this repo
• Make the TESTPYPI_API_TOKEN org secret visible to this repo
• Decide whether the repo goes public at the same time
• Name the first version published from here (no downward reset —
  audioif is past v0.1.1)

audiocomponents

• Add prepare-release.yml, tag-release.yml, publish-release-packages.yml
  (two pure-python jobs, working-directory: lib/<package>,
  mip-profile: audioinstruments,audioeffects)
• Add root VERSION; point both pyproject.toml project.urls here
• Move AUDIOIF_PIN to a real dependency floor in both pyprojects
  (pydevices-audioif>=X.Y) — the pin stays for gates

audioif

• Delete lib/audioinstruments/ and lib/audioeffects/ (needs its own approval)
• publish-release-packages.yml: drop publish-audioinstruments,
  publish-audioeffects, and mip-profile
• test-cpython.yml: drop the pip install ./lib/... step and the four
  component tests it covers
• manifest.py: drop the two freeze entries
• README / AGENTS.md / pyproject.toml comments: the lib/ description
• tests/parity/capture_render_reference.py now needs the components
  installed rather than in-tree

mip

• pydevices-lock.json: repository → PyDevices/audiocomponents, new ref
• Tag first, lockfile second — every mip run re-clones every locked
  source at its ref, so the tag must exist and already contain lib/<package>

micropython-vst3

• src/plugin/CMakeLists.txt MPVST_AUDIOIF_LIB (hard ../audioif/lib,
  staged into the plug-in bundle) → retarget, and consider renaming the variable
• scripts/bootstrap.sh: the TestPyPI install line
• Rebuild and reinstall the sidecar plug-in; re-run the render acceptance

dotgithub / anchor docs

• data/repos_db.json: new entry; regenerate the four generated sites
• audioif's own db entry: its description names the components
• anchor docs/accuracy-vision.md; the parity-verifier, song-writer and
  release-verifier agent charters carry component paths

[bdbarnett]

D1 decided at Gate 1 (2026-09-01): publishing deferred — audioif keeps shipping both packages; this checklist stays parked for a later phase boundary. The six Phase 1 remainder gates were approved and blessed the same day.

[bdbarnett]

Worth re-framing this. It was filed as a tidiness question; it is now the
reason the accuracy work does not reach anyone.

Measured today:

• 57 files differ between audioif/lib and audiocomponents/lib — most
  of the instrument library. Nothing exists only in audiocomponents, so this
  is drift, not new structure.
• audioif's copy is missing the METAL bank rebuild (_METAL_CYM_HP_HZ,
  _METAL_SEED), press_voice, and today's tr808 rimshot is ~22 dB below its kit median #9/_support.trigger_voice admits by Note count, so multi-Note voices silently cost two slots #10/tr808 cymbal is brighter than the machine after the METAL rebuild #11 fixes. logmap is in
  both, so the split is not total.
• audioif is PUBLIC and holds the publishing workflows
  (publish-release-packages.yml, tag-release.yml, prepare-release.yml).
  audiocomponents is PRIVATE and has only lint.yml and tests.yml.

So every accuracy fix since the seed lives only in the private repo that
cannot publish, while the public one ships the pre-accuracy instruments. The
kits queued for the batched listen are not the kits a user installing the
package would hear.

Not proposing a direction — which repo becomes canonical is your call, and
either answer resolves it. Flagging only that the cost of leaving it is no
longer zero: the gap grows with each fix, and a 57-file reconciliation is
easier now than after Phase 2.

No action taken. Syncing the two is not something I should do unilaterally.

[bdbarnett]

Decided (Brad, 2026-09-03). audiocomponents becomes the public, canonical
repo, as a follow-up to the current accuracy work. The old components in
audioif will be scrapped rather than reconciled.

That settles the direction of the 57-file drift measured above: it is not a
sync problem. audioif's copies are being retired, so nothing needs merging
back — the drift only mattered while both were candidates to be canonical.

Sequencing note: this is a follow-up, so the batched listen still happens
against audiocomponents as it stands. Nothing in the flip needs to land before
the ears.

Not started. Left open pending the follow-up.

[bdbarnett]

Replay report (post-tag series, 2026-09-03; local commits 4521f39 → c19818b → 5c58ec0, nothing pushed). Full text, so it is where Brad can read it:

audiocomponents replay report — audioif af837de onto the shipping copy (audiocomponents#2)

Date: 2026-09-03. Series of local commits on main, nothing pushed, no tags.

Provenance

• venv: /tmp/claude-1000/-home-brad-gh-pydevices/2825dfbc-8269-4769-8e80-f7e2e1a461a1/scratchpad/venv-v020 (python 3.12.3)
• installed pydevices-audioif direct_url.json vcs_info.commit_id: ec3c67ffc42c1bf050b7179dc55fdba28ac4bb5a (requested_revision v0.2.0)
• git -C audioif rev-parse v0.2.0^{commit}: ec3c67ffc42c1bf050b7179dc55fdba28ac4bb5a
• match: yes. Components installed --no-deps -e lib/audioinstruments -e lib/audioeffects from this tree.
• MPVST_COMPONENTS_LIB=/home/brad/gh/pydevices/audiocomponents/lib; derive_patches run from micropython-vst3 aace838 with PYTHONPATH=lib, tie-aware (9271ef6), never --force.
• cmods/bin/micropython: built 2026-09-02 01:16, predates audioif afb6b71 — STALE. Its gate leg is labelled as such and is not evidence.

Commits

• 4521f39 pin: move AUDIOIF_PIN to v0.2.0, raise both floors to match, --no-deps on the gates' editable installs
• c19818b parity: carry audioif's 79ae343 retirement and f7a370a re-captures across; fix run_instruments_parity discards the caller's PYTHONPATH, so engine-substitution tests silently no-op #25; bring _vstaudio_shim.py
• (this step) instruments: replay audioif af837de — see git log

Macros re-mapped (the elif data0 == N: lines that gained logmap), per file

andromeda         5: 9 Amp Attack (a_a), 10 Amp Decay (a_d), 12 Amp Release (a_r), 13 Filter Attack (f_a), 14 Filter Decay (f_d)
arp2600           7: 7 Env 1 Attack (e1_a), 8 Env 1 Release (e1_r), 9 Env 2 Attack (e2_a), 10 Env 2 Decay (e2_d), 12 Env 2 Release (e2_r), 13 VCA Attack (vca_a), 14 VCA Release (vca_r)
b3                1: 6 Perc Decay (perc_dec)
clavinet          3: 6 Amp Attack (amp_a), 7 Amp Decay (amp_d), 9 Amp Release (amp_r)
cp70              1: 6 Decay (decay)
cs80              6: 3 HPF Cutoff (hpf_cutoff), 8 Amp Attack (amp_a), 9 Amp Decay (amp_d), 11 Amp Release (amp_r), 12 Filter Attack (filt_a), 13 Filter Decay (filt_d)
cz101             5: 2 PD Attack (pd_a), 3 PD Decay (pd_d), 7 Amp Attack (amp_a), 8 Amp Decay (amp_d), 10 Amp Release (amp_r)
d50               3: 7 Amp Attack (amp_a), 8 Amp Decay (amp_d), 10 Amp Release (amp_r)
dx7               5: 4 Env 1 Decay (e1_d), 5 Env 2 Decay (e2_d), 6 Env 3 Decay (e3_d), 7 Release Time (rel_t), 9 Attack Time (att_t)
emulator2         3: 3 Attack (amp_a), 4 Decay (amp_d), 6 Release (amp_r)
fairlight         3: 3 Attack (amp_a), 4 Decay (amp_d), 6 Release (amp_r)
fs1r              3: 6 Amp Attack (amp_a), 7 Amp Decay (amp_d), 9 Amp Release (amp_r)
jp8000            6: 7 Amp Attack (a_a), 8 Amp Decay (a_d), 10 Amp Release (a_r), 11 Filter Attack (f_a), 12 Filter Decay (f_d), 14 HPF Cutoff (hpf)
juno106           5: 7 HPF (hpf_cutoff), 11 Amp Attack (amp_a), 12 Amp Decay (amp_d), 14 Amp Release (amp_r), 15 Filter Decay (filt_d)
jupiter8          6: 7 Amp Release (release_time), 8 Amp Attack (amp_a), 9 Amp Decay (amp_d), 11 Filter Attack (filt_a), 12 Filter Decay (filt_d), 14 HPF Cutoff (hpf_cutoff)
k2600             3: 8 Amp Attack (a_a), 9 Amp Decay (a_d), 11 Amp Release (a_r)
karplus           1: 5 Decay (decay_time)
mellotron         3: 1 Tone (tone), 4 Attack (att), 5 Release (rel)
microwave         4: 6 Filter Attack (f_a), 7 Amp Attack (amp_a), 8 Amp Decay (amp_d), 10 Amp Release (amp_r)
ms20              5: 7 EG2 Attack (eg2_a), 8 EG2 Decay (eg2_d), 10 EG2 Release (eg2_r), 13 VCA Attack (vca_a), 14 VCA Release (vca_r)
ms2000            3: 7 Amp Attack (amp_a), 8 Amp Decay (amp_d), 10 Amp Release (amp_r)
music_easel       3: 5 LPG Decay (lpg_decay), 6 Amp Attack (amp_a), 8 Amp Release (amp_r)
nord_lead         5: 9 Amp Attack (amp_a), 10 Amp Decay (amp_d), 12 Amp Release (amp_r), 13 Filter Attack (filt_a), 14 Filter Decay (filt_d)
obxa              4: 5 Brass Attack (brass_a), 6 Brass Decay (brass_d), 7 Amp Attack (amp_a), 9 Amp Release (amp_r)
odyssey           4: 8 Attack (amp_a), 9 Decay (amp_d), 11 Release (amp_r), 12 HPF Cutoff (hpf_cutoff)
pianet            3: 3 Decay Time (decay_time), 6 Amp Attack (amp_a), 8 Amp Release (amp_r)
polysix           3: 6 Amp Attack (amp_a), 7 Amp Decay (amp_d), 9 Amp Release (amp_r)
ppg_wave          4: 5 Filter Attack (f_a), 6 Filter Decay (f_d), 7 Amp Attack (amp_a), 9 Amp Release (amp_r)
prophet5          5: 7 Release (release_time), 8 Amp Attack (amp_a), 9 Amp Decay (amp_d), 11 Filter Attack (filt_a), 12 Filter Decay (filt_d)
prophet_vs        6: 7 Filter Attack (f_a), 8 Filter Decay (f_d), 10 Filter Release (f_r), 11 Amp Attack (a_a), 12 Amp Decay (a_d), 14 Amp Release (a_r)
rhodes            4: 6 Tone (tone), 7 Amp Attack (amp_a), 8 Amp Decay (amp_d), 10 Amp Release (amp_r)
sh101             1: 6 Fast Decay (fast_decay)
solina            2: 5 Attack (att), 6 Release (rel)
taurus            3: 7 Amp Attack (amp_a), 8 Amp Decay (amp_d), 10 Amp Release (amp_r)
tb303             1: 5 Decay (decay_time)
virus             3: 7 Amp Attack (amp_a), 8 Amp Decay (amp_d), 10 Amp Release (amp_r)
vl1               3: 5 Attack (amp_a), 6 Decay (amp_d), 8 Release (amp_r)
vp330             2: 4 Attack (amp_a), 5 Release (amp_r)
wasp              3: 5 Amp Attack (amp_a), 6 Amp Decay (amp_d), 8 Amp Release (amp_r)
wurlitzer         3: 5 Amp Attack (amp_a), 6 Amp Decay (amp_d), 8 Amp Release (amp_r)
TOTAL elif data0 == N lines that gained logmap: 143 across 40 files

40 files also had their patch 0 line rewritten by the patch (the values af837de derived).

derive_patches — the 40, through pydevices-audioif ec3c67ff

Run A, after apply / before derive (report-only):

audiocomponents/lib/audioinstruments/andromeda.py    16 macros
audiocomponents/lib/audioinstruments/arp2600.py      16 macros
audiocomponents/lib/audioinstruments/b3.py           11 macros
audiocomponents/lib/audioinstruments/clavinet.py     11 macros
audiocomponents/lib/audioinstruments/cp70.py          9 macros
audiocomponents/lib/audioinstruments/cs80.py         16 macros
audiocomponents/lib/audioinstruments/cz101.py        12 macros
audiocomponents/lib/audioinstruments/d50.py          12 macros
audiocomponents/lib/audioinstruments/dx7.py          16 macros
audiocomponents/lib/audioinstruments/emulator2.py    11 macros
audiocomponents/lib/audioinstruments/fairlight.py    10 macros
audiocomponents/lib/audioinstruments/fs1r.py         11 macros
audiocomponents/lib/audioinstruments/jp8000.py       16 macros
audiocomponents/lib/audioinstruments/juno106.py      16 macros
audiocomponents/lib/audioinstruments/jupiter8.py     16 macros
audiocomponents/lib/audioinstruments/k2600.py        16 macros
audiocomponents/lib/audioinstruments/karplus.py       7 macros
audiocomponents/lib/audioinstruments/mellotron.py     8 macros
audiocomponents/lib/audioinstruments/microwave.py    12 macros
audiocomponents/lib/audioinstruments/ms20.py         16 macros
audiocomponents/lib/audioinstruments/ms2000.py       12 macros
audiocomponents/lib/audioinstruments/music_easel.py  10 macros
audiocomponents/lib/audioinstruments/nord_lead.py    16 macros
audiocomponents/lib/audioinstruments/obxa.py         11 macros
audiocomponents/lib/audioinstruments/odyssey.py      16 macros
audiocomponents/lib/audioinstruments/pianet.py       10 macros
audiocomponents/lib/audioinstruments/polysix.py      11 macros
audiocomponents/lib/audioinstruments/ppg_wave.py     12 macros
audiocomponents/lib/audioinstruments/prophet5.py     16 macros
audiocomponents/lib/audioinstruments/prophet_vs.py   16 macros
audiocomponents/lib/audioinstruments/rhodes.py       13 macros
audiocomponents/lib/audioinstruments/sh101.py         8 macros
audiocomponents/lib/audioinstruments/solina.py        9 macros
audiocomponents/lib/audioinstruments/taurus.py       12 macros
audiocomponents/lib/audioinstruments/tb303.py         9 macros
audiocomponents/lib/audioinstruments/virus.py        12 macros
audiocomponents/lib/audioinstruments/vl1.py          10 macros
audiocomponents/lib/audioinstruments/vp330.py        12 macros
audiocomponents/lib/audioinstruments/wasp.py         10 macros
audiocomponents/lib/audioinstruments/wurlitzer.py    10 macros
soundtrack/Perihelion/instruments/vp330.py           SKIP no MACRO_LABELS

41 files, 0 written, 1 skipped, 0 macros needing attention

Run B, --write WITHOUT --force:

audiocomponents/lib/audioinstruments/andromeda.py    16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/arp2600.py      16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/b3.py           11 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/clavinet.py     11 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/cp70.py          9 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/cs80.py         16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/cz101.py        12 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/d50.py          12 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/dx7.py          16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/emulator2.py    11 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/fairlight.py    10 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/fs1r.py         11 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/jp8000.py       16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/juno106.py      16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/jupiter8.py     16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/k2600.py        16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/karplus.py       7 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/mellotron.py     8 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/microwave.py    12 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/ms20.py         16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/ms2000.py       12 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/music_easel.py  10 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/nord_lead.py    16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/obxa.py         11 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/odyssey.py      16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/pianet.py       10 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/polysix.py      11 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/ppg_wave.py     12 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/prophet5.py     16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/prophet_vs.py   16 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/rhodes.py       13 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/sh101.py         8 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/solina.py        9 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/taurus.py       12 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/tb303.py         9 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/virus.py        12 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/vl1.py          10 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/vp330.py        12 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/wasp.py         10 macros  (already has a patch 0)
audiocomponents/lib/audioinstruments/wurlitzer.py    10 macros  (already has a patch 0)
soundtrack/Perihelion/instruments/vp330.py           SKIP no MACRO_LABELS

41 files, 0 written, 41 skipped, 0 macros needing attention

Run C, after derive (report-only): byte-identical to run A (diff empty).

	! (unreachable/approx)	~ (drift vs committed)	SKIP
after apply, before derive	0	0	1 (soundtrack/Perihelion/instruments/vp330.py shim — stem collision with the name filter; no MACRO_LABELS, nothing written)
after derive	0	0	1 (same)

The micropython-vst3 tree stayed clean throughout (git status --porcelain empty).

Postcondition 1 — drum file sha256

Baseline taken at 7befee5; re-taken at ffed304 and c19818b (identical both times) before the patch was applied.

kit	before (7befee5 baseline)	after replay
cr78	b7c9508338ca95bb9e831aa757b6c25f8ed697096d923ca404bd7c2381686df4	b7c9508338ca95bb9e831aa757b6c25f8ed697096d923ca404bd7c2381686df4	identical
tr808	55d399f910e31ff241587d4fa68a986da36a2e00e25111c7272a908aae92ac24	55d399f910e31ff241587d4fa68a986da36a2e00e25111c7272a908aae92ac24	identical
tr909	c0fe7755e1ca19bae29a305f06ab2d0ae3ff33e26dd4e30a8b191ce15f80a835	c0fe7755e1ca19bae29a305f06ab2d0ae3ff33e26dd4e30a8b191ce15f80a835	identical
drumtraks	77cd685729508bbe98ea0a8ac54a406f73c008bb260761d694c30a54463f55e6	77cd685729508bbe98ea0a8ac54a406f73c008bb260761d694c30a54463f55e6	identical
linndrum	354104359476bd0acef49ec9affd3de3b27b5d132685f45ca449746d1fa27bd4	354104359476bd0acef49ec9affd3de3b27b5d132685f45ca449746d1fa27bd4	identical
sp1200	fa05d5f9df13432e239e674fedac2668f13f9985b35a0209fe17735f162fab4c	fa05d5f9df13432e239e674fedac2668f13f9985b35a0209fe17735f162fab4c	identical
simmons_sdsv	196b8d8829851e64dffa4c209dddbd79753a87b5cd9a6bb156082120cd1e18e3	196b8d8829851e64dffa4c209dddbd79753a87b5cd9a6bb156082120cd1e18e3	identical
dmx	00f646a043f982cbf1ba56333b669baae6a5674abfb38fd08f66fb0632247689	00f646a043f982cbf1ba56333b669baae6a5674abfb38fd08f66fb0632247689	identical
tr707	9b55b0f9299f189390879ef97589f7fbd03b1fce1ed8a925d3e87091d0dc2b5f	9b55b0f9299f189390879ef97589f7fbd03b1fce1ed8a925d3e87091d0dc2b5f	identical
tr606	e1ad92bb63517d07ffbb5d732192f4f5a490d83ac5f20b482372379b63a69f1a	e1ad92bb63517d07ffbb5d732192f4f5a490d83ac5f20b482372379b63a69f1a	identical

Postcondition 2 — rendered drum digests through the gate's own renderer

Both runs through the SAME installed core: pydevices-audioif direct_url.json commit_id ec3c67ffc42c1bf050b7179dc55fdba28ac4bb5a (venv-v020). A digest difference could therefore only have meant the replay, never the pin move. The FAILs are against the pre-rewrite ac87f13 oracle and are expected; the actual column (left digest) is what is compared.

BEFORE (tree c19818b, patch not yet applied):

# rendered-drum digests BEFORE applying af837de-instruments.patch
# tree: audiocomponents c19818b6373b9550113f933bd4b566a775af1ca2
# core: installed pydevices-audioif direct_url.json commit_id = ec3c67ffc42c1bf050b7179dc55fdba28ac4bb5a  (venv-v020)
# command: /tmp/claude-1000/-home-brad-gh-pydevices/2825dfbc-8269-4769-8e80-f7e2e1a461a1/scratchpad/venv-v020/bin/python tests/parity/run_instruments_parity.py --verify --batch drums --interpreters cpython --require-interpreters --include-rebuilt
interpreters: cpython

FAIL     cr78           cpython        2b319029a52bfa30 != 9f30aaf6b67ceca1
FAIL     dmx            cpython        9319fb93e57d1ed2 != 68c8ba056eaf1b23
FAIL     drumtraks      cpython        af887a1660d52911 != c67e86902a07eba5
FAIL     linndrum       cpython        403607f0d6019019 != 109e4dc28d5d9304
FAIL     simmons_sdsv   cpython        d2e922e4e52c5721 != dfe65e9b69b83fe7
FAIL     sp1200         cpython        105a93c3ff16732d != a177b1ab4cf05767
FAIL     tr606          cpython        eedce4aa1dbeb667 != 787c90daee0e09c2
FAIL     tr707          cpython        abb5afad7da08092 != 8f06eab1dc115fcd
FAIL     tr808          cpython        4faac76f57241576 != 676dc91c01c72c45
FAIL     tr909          cpython        0c756243760ed4ab != 5e9a3bd327127e24

10 comparisons, 10 failures, 0 rebuilt (not compared)
  cr78 on cpython
  dmx on cpython
  drumtraks on cpython
  linndrum on cpython
  simmons_sdsv on cpython
  sp1200 on cpython
  tr606 on cpython
  tr707 on cpython
  tr808 on cpython
  tr909 on cpython

AFTER (replay applied, derive run, file sha256 asserted):

# rendered-drum digests AFTER af837de replay + derive + file-sha256 assertion
# tree: audiocomponents c19818b6373b9550113f933bd4b566a775af1ca2 + staged replay of the 40 (uncommitted at run time)
# core: installed pydevices-audioif direct_url.json commit_id = ec3c67ffc42c1bf050b7179dc55fdba28ac4bb5a  (venv-v020)
# command: /tmp/claude-1000/-home-brad-gh-pydevices/2825dfbc-8269-4769-8e80-f7e2e1a461a1/scratchpad/venv-v020/bin/python tests/parity/run_instruments_parity.py --verify --batch drums --interpreters cpython --require-interpreters --include-rebuilt
interpreters: cpython

FAIL     cr78           cpython        2b319029a52bfa30 != 9f30aaf6b67ceca1
FAIL     dmx            cpython        9319fb93e57d1ed2 != 68c8ba056eaf1b23
FAIL     drumtraks      cpython        af887a1660d52911 != c67e86902a07eba5
FAIL     linndrum       cpython        403607f0d6019019 != 109e4dc28d5d9304
FAIL     simmons_sdsv   cpython        d2e922e4e52c5721 != dfe65e9b69b83fe7
FAIL     sp1200         cpython        105a93c3ff16732d != a177b1ab4cf05767
FAIL     tr606          cpython        eedce4aa1dbeb667 != 787c90daee0e09c2
FAIL     tr707          cpython        abb5afad7da08092 != 8f06eab1dc115fcd
FAIL     tr808          cpython        4faac76f57241576 != 676dc91c01c72c45
FAIL     tr909          cpython        0c756243760ed4ab != 5e9a3bd327127e24

10 comparisons, 10 failures, 0 rebuilt (not compared)
  cr78 on cpython
  dmx on cpython
  drumtraks on cpython
  linndrum on cpython
  simmons_sdsv on cpython
  sp1200 on cpython
  tr606 on cpython
  tr707 on cpython
  tr808 on cpython
  tr909 on cpython

Line-for-line comparison of the ten per-kit lines: identical.

The runner prints 16 hex chars; the full 64-hex digests via the runner's own probe (instruments_probe_new.py, same cwd and PYTHONPATH) are also identical:

cr78           2b319029a52bfa3071d59e9fe6653ddf21afd6bd87e8aba356863a97a49338e8
dmx            9319fb93e57d1ed23d6e632a8a75f320ff3d261cdd8fcccc882f8d93534ebd45
drumtraks      af887a1660d52911b25f4a15453b90ccb1b6b82cedaaf009cb4c88fd843e5c0d
linndrum       403607f0d60190191ba6a0d0be40d0b860c01b364d28dc141db86314a2b1baca
simmons_sdsv   d2e922e4e52c572193435c1b9983109029b76a809237688ef2b2d09aac8a793e
sp1200         105a93c3ff16732d315a938450eba26d555dfcf12b66fe8b449b9313dda4f001
tr606          eedce4aa1dbeb667029b572610cae68eace9d23bd77feedd5b6242ded604a4ec
tr707          abb5afad7da08092d32c6099863a13e39ff539314defff7070f8281bdfc39a0b
tr808          4faac76f57241576d6b24f725789870d89cf1d6406fe68b3113910c58e8106cc
tr909          0c756243760ed4ab551f41eea521e3138c9db63ea6722890ec9783ada0585ced

Refused presses (FINDING)

Synthesizer.press in the CPython twin refuses silently when the engine holds 14 notes and none is released (if victim is None: continue, no counter, nothing on stderr). Counted by wrapping press with that exact predicate and re-running the gate's probe over all 93 modules in the same env:

• all ten drum kits, cr78 and tr808 included: 0 refused (73–186 presses each).
• 11 melodic modules refuse under the parity sequence: solina 22, jp8000 14, virus 14, b3 10, vp330 10, farfisa 6, k2600 2, polysix 2, ppg_wave 2, prophet_vs 2, vox_continental 2 — 86 refusals in 8672 Note presses; no int presses anywhere.
• The same eleven measured against the pre-replay tree (c19818b, via git archive): identical counts, every one. farfisa and vox_continental are outside the 40. The refusals belong to the sequence and the 14-Note cap, not to af837de; they are reported here because a refused press is a finding, not because the replay caused them.

Full per-module table: /tmp/claude-1000/-home-brad-gh-pydevices/2825dfbc-8269-4769-8e80-f7e2e1a461a1/scratchpad/refusals.txt; pre-replay: /tmp/claude-1000/-home-brad-gh-pydevices/2825dfbc-8269-4769-8e80-f7e2e1a461a1/scratchpad/refusals-pre-replay.txt.

Anomalies / surprises

1. carry-across.patch DROPPED the ten Phase 1 drum entries from REBUILT while adding the 40 (its hunk replaced the dict body). replay-task.md §1 says to KEEP them and the expected count was 40+10, so they were restored by hand after applying; 50 entries committed in c19818b.
2. The run_instruments_parity discards the caller's PYTHONPATH, so engine-substitution tests silently no-op #25 acceptance as literally specified (--batch drums, no --include-rebuilt) is vacuous once all ten kits are retired: 0 comparisons, 0 failures, 10 rebuilt, rc 0, no probe runs. With --include-rebuilt the planted synthio.py fires at the first probe (ImportError: planted, rc 1) — the fix took.
3. git apply -3 stages the hunks; the per-file listing had to be read from git diff --cached.
4. vp330 is both one of the 40 and a soundtrack shim stem; derive's name filter picks the shim up too (SKIP, nothing written).
5. flake8 is not installed on the system or in the venv; see the gate section below for the throwaway-venv run.

Gate after the replay commit

Replay commit: 5c58ec0. All batches, provenance venv (core ec3c67ff).

run	command	summary line	FAIL lines
step 3 (at the tag, before replay)	--interpreters cpython,micropython --require-interpreters	86 comparisons, 0 failures, 50 rebuilt (not compared)	none
step 3, cpython leg	--interpreters cpython	43 comparisons, 0 failures, 50 rebuilt (not compared)	none
step 3, micropython leg — STALE BINARY (2026-09-02 01:16, predates afb6b71), to be re-run after rebuild, not evidence	--interpreters micropython	43 comparisons, 0 failures, 50 rebuilt (not compared)	none
step 4 (after replay)	--interpreters cpython,micropython --require-interpreters	86 comparisons, 0 failures, 50 rebuilt (not compared)	none
step 4, cpython leg	--interpreters cpython	43 comparisons, 0 failures, 50 rebuilt (not compared)	none
step 4, micropython leg — STALE BINARY, same label	--interpreters micropython	43 comparisons, 0 failures, 50 rebuilt (not compared)	none

The 40 all report rebuilt ... af837de - time and filter macros log-mapped; the ten drums report rebuilt ... phase 1; every ok/rebuilt line is identical between the step-3 and step-4 runs on both legs (diff empty). Full outputs: gate3-*.txt, gate4-*.txt in this directory.

flake8 7.3.0 with the repo's .flake8 (F,E9,W6) over the runner, the shim and the 40: clean (throwaway venv-lint).

tests.yml steps in the provenance venv (step 1)

• unittest discover: Ran 112 tests, OK (skipped=1)
• tools/validate_api.py: validated 53 instruments and 46 effects
• tests/parity/effects_library_smoke.py: 46 classes, 47 patches, 0 failures

#25 acceptance text (step 2)

As specified (--batch drums --interpreters cpython --require-interpreters, planted synthio.py on PYTHONPATH): rc 0, 0 comparisons, 0 failures, 10 rebuilt (not compared) — no probe ran, all kits retired.

With --include-rebuilt: rc 1 —

Traceback (most recent call last):
  File "/home/brad/gh/pydevices/audiocomponents/tests/parity/instruments_probe_new.py", line 90, in <module>
    main()
  File "/home/brad/gh/pydevices/audiocomponents/tests/parity/instruments_probe_new.py", line 43, in main
    module = audioinstruments.load(name)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/brad/gh/pydevices/audiocomponents/lib/audioinstruments/__init__.py", line 71, in load
    __import__(full)
  File "/home/brad/gh/pydevices/audiocomponents/lib/audioinstruments/cr78.py", line 59, in <module>
    import synthio
  File "/tmp/claude-1000/-home-brad-gh-pydevices/2825dfbc-8269-4769-8e80-f7e2e1a461a1/scratchpad/plant-v020/synthio.py", line 1, in <module>
    raise ImportError("planted")
ImportError: planted
interpreters: cpython

Traceback (most recent call last):
  File "/home/brad/gh/pydevices/audiocomponents/tests/parity/run_instruments_parity.py", line 355, in <module>
    main()
  File "/home/brad/gh/pydevices/audiocomponents/tests/parity/run_instruments_parity.py", line 352, in main
    verify(args, interpreters)
  File "/home/brad/gh/pydevices/audiocomponents/tests/parity/run_instruments_parity.py", line 281, in verify
    actual = digest(run_probe(prefix, "instruments_probe_new.py",
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/brad/gh/pydevices/audiocomponents/tests/parity/run_instruments_parity.py", line 192, in run_probe
    raise RuntimeError("probe failed: %s %s" % (probe, " ".join(probe_args)))
RuntimeError: probe failed: instruments_probe_new.py cr78

[bdbarnett]

Landed, 2026-09-03. Every item of the rewiring is in:

• audioif v0.2.0 — core only (audiorender in the wheel, neither component package; no MIP profile); the retired lib/audioinstruments and lib/audioeffects deleted (3770b02); render reference re-captured against this repository at the tag (effects: Phase 0 — The survey #28 there, closed).
• audiocomponents v0.2.0 — first release from here, continuing the packages' version history (0.0.4 … 0.1.1 came from audioif; those five tags now live here too). AUDIOIF_PIN = v0.2.0 with matching floors and --no-deps on the gates' editable installs; audioif's parity retirement and re-captures carried across with the run_instruments_parity discards the caller's PYTHONPATH, so engine-substitution tests silently no-op #25 fix; the af837de log-mapping replayed against this tree with the ten drum kits proven byte- and render-identical; the three tr808 fixes; all of it heard and approved by Brad.
• MIP index — both entries repointed (mip f94ef5e9) and, after the tools' own profile map was corrected (.github 9aef8e68, pinned by mip c1d73393), both published: audioinstruments and audioeffects latest.json read 0.2.0 as of 23:31Z.
• micropython-vst3 — stages the components from here (MPVST_COMPONENTS_LIB, alias honoured for one release: Gate A: tr808 dossier awaiting APPROVE ACCURACY DOSSIER tr808 #4 there).
• Org website — audiocomponents page live, audioif page describes the core only, org profile README regenerated.

Loose ends live in their own issues, not here: #24/#25/#26 (parity artifacts, the vacuous-batch rule, the eleven refusing instruments), .github#33/#34/#35 (runbook, release-health race, cutting the publishing tag that carries the corrected map), micropython-vst3#4/#5, audioif#27/#30.