- Hand /generate ids to the global SSE watcher so playback fires on completion. The mutation onSuccess was checking audio_path on a queued row, which is always empty — autoplay never ran.
- Bind the Play As voice selection to capture_settings.default_playback_voice_id, kept in sync with the Settings → Captures and Settings → MCP pickers. Picking from the split-button dropdown writes back to settings.
- Extract AudioBars from HistoryTable into a shared component; use it for the Play As generating state in place of Loader2.
- Stop the active-state hover from flashing white text when the button is in its lighter accent/10 fill.
- Drop the gradient avatar swatches from the Settings → Captures voice dropdown.
- Backend: when the gen worker exits without writing a terminal status (e.g. SQLite lock racing the failed-status write inside its own exception handler), the cancel endpoint now flips the row to failed instead of 409-ing. Worker also force-fails on its way out as a belt-and-suspenders.
ListPane is a compound component (Header / TitleRow / Title / Actions /
Search / Scroll) that owns the relative wrapper, faded right divider
(50px top fade), top scroll mask, and absolute-positioned header used by
every list-detail tab. Wires up CapturesTab, StoryList, and EffectsList.
EffectsTab gets -mx-8 / pr-8 to match the edge-to-edge layout used
elsewhere.
Other changes:
- MCPPage: native <select> → shadcn <Select> for default voice and
per-binding voice pickers
- Button outline variant: add hover:border-accent
- Drop hover:text-destructive from trailing delete buttons
(HistoryTable, GpuAcceleration, GpuPage, EffectsChainEditor,
EffectsDetail)
- HistoryTable empty state moved behind t('history.empty')
- StoryContent scroll padding pt-14 → pt-16
- backend health reports the captures dir
- landing CapturesMockup: "Send to" → "Export" with Download icon
- CHANGELOG: drop [Unreleased] personality section
POST /speak is a REST wrapper around voicebox.speak for agents that
don't talk MCP (shell scripts, ACP, A2A). It reads X-Voicebox-Client-Id
and uses it for the same per-client profile resolution + default
personality lookup the MCP tool does (speak.py:39-64), so its callers
are first-class clients — but the ClientIdMiddleware only stamped
last_seen_at on /mcp* paths. REST speak callers showed up as "never
seen" in Settings → MCP despite actively acting on their bindings.
Widen the stamp predicate to an explicit ("/mcp", "/speak") prefix
list, and require a path boundary on match so future routes named
/mcpfoo or /speakers don't silently inherit the stamp via the prefix.
New test_client_id_middleware.py pins the scope with 17 parametrised
cases (both the allowed set and the overlap cases that must not match).
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
The inline lifespan ran _run_shutdown inside the MCP context, so the
TTS / Whisper / LLM models were unloaded *before* FastMCP's __aexit__
got a chance to cancel its in-flight session tasks. Any MCP request
mid-generate at shutdown time would crash on "model unloaded" instead
of receiving a clean session-cancelled error.
Rewire via compose_lifespan (which was already defined in
mcp_server.server for exactly this purpose but never used):
AsyncExitStack enters factories in order and exits in LIFO, so
MCP teardown fires first — cancelling sessions — and _run_shutdown
runs after nothing is holding the models. Smoke test shows the
log order flipped as expected:
Ready
StreamableHTTP session manager started
... running ...
StreamableHTTP session manager shutting down ← was last, now first
Voicebox server shutting down... ← was first, now last
As a side benefit, _run_shutdown is now paired with _run_startup via
try/finally inside voicebox_lifespan, so a partial startup (models
half-loaded, MCP __aenter__ fails) still unloads whatever was loaded
instead of leaking it to process exit.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
SQLite gained ALTER TABLE … DROP COLUMN in 3.35 (Mar 2021). Production
PyInstaller builds bundle Python 3.12 which links to SQLite 3.40+ so
that path is always safe, but a dev running the backend directly on
Ubuntu 20.04 (3.31) or Debian 11 (3.34) would crash on first startup
trying to drop the legacy default_intent column.
Add _supports_drop_column(engine) — returns True on non-SQLite
dialects (Postgres / MySQL have supported DROP COLUMN for decades) and
gates on the runtime sqlite_version for SQLite. When unsupported, log a
warning and leave the unused column in place: SQLAlchemy only maps
declared columns, so a stray default_intent column does no reads or
writes and can't interfere with runtime behaviour.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
The word-level pass catches single-word Whisper loops ("URL URL URL…")
but misses two common hallucination patterns the PR had to claim as
"edge cases":
1. Multi-word English loops — "thanks for watching thanks for watching…"
× 6 sails through because no two consecutive tokens are identical
after text.split().
2. CJK loops — "謝謝觀看" × 7 sails through because text.split() returns
a single unsplit token for the whole loop (no whitespace between
characters).
Add a character-level second pass: a non-greedy regex finds any 2–60
char substring that repeats min_run+ times immediately after itself and
strips the run. The 2-char floor keeps emphasised single-letter runs
("wooooooow") intact. The 60-char ceiling covers every observed
Whisper tail hallucination ("Please like and subscribe to my
channel.", "Subtitles by the Amara.org community") while staying short
enough that coincidental long-phrase repetition in legitimate speech
doesn't hit the threshold. Whitespace normalisation only runs when the
pass actually stripped something, so untouched transcripts keep their
original spacing.
New test_refinement_collapse.py gives the pre-processor its first
deterministic unit-test coverage: 17 tests pinning the word-level
legacy behaviour plus the new multi-word English / CJK / Japanese /
emphasis-preservation cases.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
- Collapse intent tri-state (respond/rewrite/compose) to `personality: bool` on /generate, /speak, and voicebox.speak. Drop respond entirely; keep compose as a standalone button via /profiles/{id}/compose. Remove /rewrite, /respond, and /speak profile endpoints.
- FloatingGenerateBox: Wand2 persona toggle + Dices compose button appear when the selected profile has a personality. ProfileCard badges Wand2 alongside the effects Sparkles.
- MCP bindings: default_intent column → default_personality: bool. Migration drops the legacy column.
- i18n: en / ja / zh-CN / zh-TW translation files filled out and wired through the capture, server, and profile UI.
```ts
voicebox.speak({
text: "Deploy complete.",
profile: "Morgan",
personality: true, // rewrite through the profile's personality LLM
});
```
Stops the "stuck pill" failure where pressing the chord with missing
STT/LLM models triggers a recording that has nowhere to land. The
hotkey now stays disarmed until every gate (models downloaded, Input
Monitoring + Accessibility granted) is green; the empty-state checklist
in CapturesTab surfaces each unmet gate with a one-click action and
auto-arms the chord once everything turns green.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
The pill window now surfaces for agent-initiated speech without main-window
involvement. Rust subscribes to /events/speak via a tokio task + reqwest
streaming body (speak_monitor.rs), shows the pill, and forwards events to
the dictate webview over Tauri's event bus. The pill plays audio via a
plain HTMLAudioElement and emits dictate:hide when playback ends. The
pill stays hidden through the ~1 s generation wait and only surfaces when
audio actually starts, with the counter armed at that moment.
Fixes a shared-dict mutation in mcp_server/events.publish() that caused
the second subscriber (Rust speak_monitor) to receive `event: message`
instead of named speak-start/speak-end frames. Also teaches the speak_monitor
parser to handle CRLF framing (sse-starlette default). Main-window
AudioPlayer now skips autoplay for source in {mcp, rest} to avoid
double-play when both windows are alive.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
Ships the Capture release end to end. Global-hotkey dictation with
synthetic paste into the focused app on macOS and Windows, an on-screen
pill across recording / transcribing / refining, customizable push-to-
talk and toggle chords, and an accessibility-permission prompt scoped to
Settings → Captures with inline re-check feedback.
Voice profiles gain optional personalities that power compose / rewrite /
respond actions via a local Qwen3 LLM — shared with refinement, so there
is one local LLM in the app, not two.
Refinement hardened with deterministic Whisper-loop collapse before the
LLM sees the transcript, per-capture flag snapshots for re-runs, and a
ten-transcript evaluation harness across every bundled refinement size.
Version bump 0.4.5 → 0.5.0.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
* fix(offline): patch transformers mistral-regex check to survive HF failures
transformers 4.57.x's `PreTrainedTokenizerBase._patch_mistral_regex` calls
`huggingface_hub.model_info(repo_id)` unconditionally during any non-local
tokenizer load to probe for Mistral-family models. The call raises on
`HF_HUB_OFFLINE=1`, on network outages, and on slow/blocked HF endpoints,
and transformers doesn't catch any of it — the exception bubbles out of
`from_pretrained` and kills the load for unrelated engines (Qwen TTS,
Qwen CustomVoice, TADA, etc.).
0.4.2's load-time `force_offline_if_cached` guard walked straight into
this trap: on cached online users it flipped `HF_HUB_OFFLINE=1` and
converted a healthy load into a hard crash. 0.4.3's inference-path guard
masked it; #524 removed the inference guard in 0.4.4, and users updating
to 0.4.4 started hitting the same error on the load path instead
(#526).
Fix:
- Wrap `_patch_mistral_regex` so any exception from the inner HF
metadata check is swallowed and the tokenizer is returned unchanged.
Voicebox never loads Mistral models, so the regex rewrite this check
gates is a no-op for us; matches the success-path behavior for
non-Mistral repos (tokenization_utils_base.py:2503).
- Drop the `force_offline_if_cached` wraps from every load path
(pytorch_backend Qwen + Whisper, qwen_custom_voice_backend,
mlx_backend Qwen + Whisper). With the mistral patch in place they
provide zero value and only risk re-introducing the same class of
bug. Helper and its unit tests stay — still correct for targeted
future use.
- Add `backend/tests/test_offline_patch.py` covering
OfflineModeIsEnabled / ConnectionError suppression, success
pass-through, idempotence, and the missing-method no-op path.
Fixes#526.
* fix(offline): install mistral-regex patch for non-MLX backends
The previous commit left the patch wired only through ``mlx_backend.py``'s
existing import of ``hf_offline_patch``. On Windows/Linux/CUDA users who
never load the MLX backend (everyone who hit #526), the patch module was
never imported, so ``patch_transformers_mistral_regex`` never ran and the
crash persisted.
Hoist the import into ``backends/__init__.py``. Every backend imports from
this package, so the module-level patch install runs before any
``from_pretrained`` call regardless of which engine the user picks.
Caught by CodeRabbit and Cursor Bugbot on #530.
0.4.3 wrapped every inference body (`generate`, `transcribe`,
`create_voice_clone_prompt`) with `force_offline_if_cached(True, …)` to
prevent lazy HF lookups from hanging when the network drops
mid-inference (#462). That trade broke online users: the guard flips
`huggingface_hub.constants.HF_HUB_OFFLINE` globally, so any legitimate
metadata call the library makes during generation (e.g. revision
resolution via `HfApi().model_info`) now raises:
Cannot reach https://huggingface.co/api/models/Qwen/Qwen3-TTS-…:
offline mode is enabled.
Hit by multiple users on 0.4.3 within hours of release. The offline
blast radius is much larger than the original hang it fixed.
This reverts the inference-path guards. Load-path guards stay — those
worked fine in 0.4.2 and aren't the source of the regression. The
`force_offline_if_cached` helper itself is unchanged; tests still pass.
The #462 hang (network dropping mid-inference) remains unaddressed by
this commit and will need a targeted fix that doesn't flip a global
flag — most likely per-call timeouts or library-specific
`local_files_only` arguments, not a process-wide env mutation.
fugashi (pulled in by misaki[ja]) needs a MeCab dictionary at runtime.
The `unidic` package that ships today contains no data — it relies on
`python -m unidic download` (~526MB), which isn't run by `just setup`
and won't survive PyInstaller freezing.
Switch to `unidic-lite`, which bundles a MeCab-compatible dict inside
the wheel (~50MB). Collect its data files in build_binary.py so frozen
builds also pick up the dicdir. Same failure mode and same fix shape as
the existing en_core_web_sm pre-install.
mlx-audio's STT path imports miniaudio, but we install mlx-audio
--no-deps to dodge its transformers>=5.x pin. Nothing else pulls
miniaudio transitively, so fresh Apple Silicon installs fail to
transcribe with ModuleNotFoundError: miniaudio. Listed explicitly
and updated the stale comments in requirements-mlx.txt and
release.yml that claimed it came from other engines.
Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>
* fix(offline): guard inference paths with HF_HUB_OFFLINE (#462)
PR #443 wrapped the model *load* path with `force_offline_if_cached` so
cached models don't phone home at startup. The context manager restores
`HF_HUB_OFFLINE` on exit, which left inference paths (generate,
transcribe, voice-prompt creation) unguarded — and `qwen_tts`,
`mlx_audio`, and `transformers` perform lazy tokenizer/processor/config
lookups during inference. With internet on, those lookups are
near-instant and invisible; with internet off, `requests` hangs on DNS
or connect until the network returns. This is exactly what users in
#462 describe: model shows "Loaded", internet drops, generation
"thinks" forever, internet comes back, generation completes.
Chatterbox and LuxTTS don't exhibit this because their engine libs
resolve everything through already-cached paths at load time.
Fix: wrap each inference-sync body with `force_offline_if_cached(True,
...)`. Since inference only runs after a successful load, weights are
known to be on disk, so `is_cached=True` is unconditional.
Also adds the load-time guard that was missing from
`qwen_custom_voice_backend.py` — CustomVoice previously had no offline
protection at all.
Paths patched:
- PyTorchTTSBackend.create_voice_prompt (create_voice_clone_prompt)
- PyTorchTTSBackend.generate (generate_voice_clone)
- PyTorchSTTBackend.transcribe (Whisper generate + decoder-prompt-ids)
- MLXTTSBackend.generate (mlx_audio generate, all branches)
- MLXSTTBackend.transcribe (mlx_audio whisper generate)
- QwenCustomVoiceBackend._load_model_sync + generate
Does not address the secondary `check_model_inputs() missing 'func'`
error reported in the same issue — that's a `transformers` 5.x
version-skew bug on the install path, separate concern.
Fixes#462.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
* fix(offline): mutate cached HF constants + threadsafe refcount
Review feedback on the initial fix surfaced two real issues:
1. ``os.environ`` toggles alone don't flip offline mode.
``huggingface_hub.constants.HF_HUB_OFFLINE`` is read once at import
time into a module-level bool; ``transformers.utils.hub._is_offline_mode``
mirrors that bool at its own import time. The hot paths
(``_http._default_backend_factory`` in huggingface_hub,
``is_offline_mode`` in transformers) read the cached bools — not the
env — so mutating only ``os.environ`` was a no-op.
2. Race condition on concurrent inference. Two threads running inside
``force_offline_if_cached`` via ``asyncio.to_thread`` could have
thread A's ``finally`` strip thread B's offline protection mid-run.
Rewrite the helper to:
- mutate ``huggingface_hub.constants.HF_HUB_OFFLINE`` and
``transformers.utils.hub._is_offline_mode`` directly
- refcount concurrent users under a single ``threading.RLock`` so a
shared offline window is restored only when the last caller exits
- still write ``os.environ`` for anything that reads it dynamically
Also addresses the unused-variable ruff flag on the Whisper transcribe
path (``audio, sr`` → ``audio, _sr``).
New unit tests cover the cached-constant mutation, env propagation,
no-op on ``is_cached=False``, nested contexts, and a threaded race
where a slow thread must retain offline mode after a peer exits.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
* fix(offline): atomic entry rollback + tidy test assertions
Review follow-up:
- Wrap the `_offline_refcount == 0` setup in a try/except so any failure
during the cached-constant mutation (including unexpected non-ImportError
like RuntimeError or AttributeError from a half-initialized module)
rolls back *all* partial state before re-raising. Without this, a
mid-setup crash could leave `huggingface_hub.constants.HF_HUB_OFFLINE`
mutated but the refcount at 0 — a persistent offline flag outliving
the process.
- Swap ruff-flagged Yoda comparisons in the new test file (SIM300) and
add a module-level note warning that these tests mutate global state
and are not safe under cross-process parallelism.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
* test(offline): make concurrency test deterministic and bounded
Replace the `sleep(0.15)` ordering hack with an explicit `threading.Event`
the fast thread sets in `finally`. The slow thread waits on that event
(bounded), then observes the flag — so we deterministically verify the
slow thread still sees offline mode after the fast thread has exited.
Also add timeouts to `barrier.wait()` and assert `not thread.is_alive()`
after the joins so the test can't hang on an unexpected failure path.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>
* fix(audio): preprocess reference samples instead of rejecting them
Uploaded/recorded voice samples were rejected outright whenever the peak
exceeded 0.99 ("Audio is clipping (reduce input gain)"). That wasn't
actionable: a recording in the app has no pre-gain control, and an
already-captured file can't be re-taken by the user. The Settings
"Normalize audio" toggle only affects generated TTS output, so users who
enabled it expecting it to help with sample uploads were still blocked.
Replace the hard reject with a small, always-on preprocess step that
runs right after load:
- DC-offset removal
- Conservative edge-silence trim (top_db=30) with 100 ms padding kept
- Peak cap at 0.95 if the input peak exceeds that
Duration and RMS checks now run on the preprocessed waveform, so
samples that were previously rejected for being "hot" are accepted and
stored with safe headroom. True in-waveform clipping artifacts still
can't be repaired — peak scaling only prevents downstream re-clipping
during multi-sample combination and TTS inference.
Adds a unit-test file (previously none existed for audio.py).
Fixes#456.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
* fix(audio): raise trim threshold, cap pad at net-neutral
Review feedback on the preprocessor:
1. ``trim_top_db=30`` was labelled "conservative" in the docstring but is
actually *more* aggressive than librosa's default of 60. Normal
speech dynamic range sits around 30 dB, so 30 dB would eat quiet
trailing syllables and soft consonants. Raise the default to 40 dB —
below normal speech dynamic range but still catching obvious edge
silence — and fix the docstring.
2. Unconditional 100 ms edge padding ran even when ``librosa.effects.trim``
removed nothing. For a well-recorded 29.9 s upload that path would
push the waveform past the 30 s ceiling and trigger a spurious "too
long" rejection. Only pad when trimming actually shortened the
audio, and cap the pad so the output never exceeds the input length.
Adds a regression test for the net-neutral length behaviour.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>
The 0.6B slot was aliased to the 1.7B repo as a temporary fallback
because `mlx-community/Qwen3-TTS-12Hz-0.6B-Base-bf16` wasn't published
when MLX support shipped. That conversion is live now, so use it —
Apple Silicon users picking 0.6B get the actual 0.6B model (1.2 GB
instead of 3.5 GB).
Also drops the now-obsolete troubleshooting entry and updates the
triage notes in PROJECT_STATUS.md.
Fixes#485.
Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>
The previous fix (#481) capped transformers at 4.57.6 in requirements-mlx.txt,
but pip's clean resolver in CI can't satisfy that alongside mlx-audio>=0.3.1
(declares `transformers==5.0.0rc3` or `>=5.0.0`) — it backtracks through every
transformers and tokenizers version and exits with `ResolutionImpossible`.
The dev install worked only because mlx-audio 0.4.1 was already present, so
pip never tried to re-resolve.
mlx-audio 0.4.1 + mlx-lm 0.31.1 both declare transformers>=5.x but the API
surface we actually use works fine on 4.57.x in practice (verified across all
engines in dev). Install both --no-deps to bypass the resolver; transitive
runtime deps (huggingface_hub, librosa, numpy, numba, pyloudnorm, etc.) are
already pulled in by requirements.txt.
Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>
mlx-audio depends on `transformers` with no upper bound. Installing
requirements-mlx.txt after requirements.txt lets pip upgrade transformers
past the 4.57.x cap to 5.x, which breaks three engines in the frozen MLX
bundle:
- qwen-custom-voice: `check_model_inputs` was rewritten to take `func` as
positional, so `@check_model_inputs()` factory calls fail with
`TypeError: missing 1 required positional argument: 'func'`
- tada-1b: `PretrainedConfig.__init_subclass__` now applies `@dataclass`,
which rejects tada's `strides: list = []` mutable default
- luxtts: Whisper init hits `AssertionError` in `torch._refs.normal_`
Restating the same constraint here keeps mlx-audio's transformers
dependency from quietly winning the resolver.
Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>
When migrating model path with an empty cache, backend returned early without emitting migration completion SSE, causing frontend overlay to hang. This patch emits complete status for empty migrations.
* fix(build): bundle kokoro source files for transformers runtime introspection
transformers opens .py source files at runtime to check attention/MoE
implementation via regex (e.g. _can_set_attn_implementation). PyInstaller's
--hidden-import only bundles .pyc bytecode, so kokoro/modules.py was missing
from the bundle causing a FileNotFoundError on Kokoro model load.
Switch from individual --hidden-import entries to --collect-all kokoro in both
build_binary.py and voicebox-server.spec. The kokoro package is 172K so no
meaningful bundle size impact.
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
* fix(build): use SPECPATH for runtime hook instead of hardcoded absolute path
The linter expanded runtime_hooks=[] to an absolute /Users/... path which
would break CI and other dev machines. Use os.path.join(SPECPATH, ...) to
mirror the relative approach in build_binary.py.
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
* fix(build): runtime hook to work around PyInstaller + Python 3.12 import breakages
Four distinct bundling-specific crashes blocked Kokoro and Qwen CustomVoice
from loading in the frozen binary:
1. torch._dynamo import triggered via class-body decorators
(@torch._dynamo.allow_in_graph on PreTrainedModel,
@torch.compiler.disable in flex_attention) pulls in torch._numpy._ufuncs
which crashes on module load with NameError: name 'name' is not defined.
2. AlbertModel (Kokoro) triggers @auto_docstring -> modeling_auto ->
GenerationMixin -> candidate_generator -> sklearn -> scipy, which hits
the same class of bug in scipy.stats._distn_infrastructure (NameError:
name 'obj' is not defined).
3. AutoModel (Qwen) pulls the same sklearn -> scipy chain directly.
4. librosa (required by most TTS engines) -> scipy.signal -> scipy.stats
hits the _distn_infrastructure crash regardless of the transformers
stubs above.
The root cause of (1) and (4) is that PyInstaller's frozen importer runs
module-level `for X in [<list-comp using dir()>]:` loops with an empty
iterable, leaving the loop variable unbound. Trailing `del obj` / unrelated
references then crash.
Fix: a single runtime hook (pyi_rth_torch_compiler_disable.py) installs:
- sys.modules stubs for torch._dynamo and torch._dynamo.config, plus a
meta-path finder for torch._dynamo.* submodules — voicebox never uses
torch.compile/dynamo for inference, so a permissive no-op stub (callable
as decorator, falsey as predicate, context-manager-safe for
TransformGetItemToIndex) is drop-in safe.
- meta-path finder stubs for transformers.utils.auto_docstring and
transformers.generation.candidate_generator — both import-chain
short-circuits; docstrings and speculative decoding aren't used for TTS.
- meta-path finder for scipy.stats._distn_infrastructure that reads the
real .py source via the wrapped loader's get_source(), replaces the
bundling-broken `del obj` with `globals().pop('obj', None)`, and
compile+exec's the patched source. This keeps the real scipy module
intact so librosa and everything downstream works normally.
Supporting changes:
- backend/pyi_hooks/hook-scipy.stats._distn_infrastructure.py sets
module_collection_mode = "pyz+py" so the .py source is actually in the
bundle for the runtime patcher to read.
- build_binary.py and voicebox-server.spec register the runtime hook and
the new hooks dir.
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
* fix(build): force transformers torch<2.6 mask path and bundle spacy_pkuseg
- patch transformers.masking_utils to set _is_torch_greater_or_equal_than_2_6
= False, forcing sdpa_mask_older_torch and avoiding the vmap .item() crash
that breaks Qwen CustomVoice generation (our torch._dynamo stub can't
reproduce TransformGetItemToIndex's graph transform).
- add PyInstaller hook to bundle transformers.masking_utils .py source so the
runtime finder can source-patch it.
- --collect-all spacy_pkuseg so Chatterbox Multilingual can load its Chinese
segmenter (dicts/default.pkl + native .so extensions).
- add per-finder install diagnostics + _HOOK_VERSION marker to make future
bundle-only regressions easier to triage.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
* fix(build): pass PyInstaller hook paths relative so .spec is portable
Absolute paths ended up in the auto-regenerated voicebox-server.spec
because build_binary.py prefixed every --runtime-hook and
--additional-hooks-dir with str(backend_dir / ...). That broke builds
on any machine whose checkout wasn't at /Users/jamie/... and anyone
invoking pyinstaller voicebox-server.spec directly.
os.chdir(backend_dir) already runs before PyInstaller (same reason
server.py works as a bare filename), so the backend_dir prefix is
unnecessary. Drop it so the generated spec references pyi_hooks/,
pyi_rth_numpy_compat.py, pyi_rth_torch_compiler_disable.py as repo-
relative paths.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <[email protected]>
0.3.0 sometimes stored relative media paths with the data-dir name baked in
(e.g. "data/profiles/<uuid>/sample.wav"). resolve_storage_path joined those
directly with _data_dir, producing "<data_dir>/data/profiles/..." — a
spurious double nest that breaks file reads after upgrading to 0.4.0.
The 0.4.0 startup migration didn't catch it because resolve_storage_path
produced the buggy double-nested path, to_storage_path saw "data" at the
first (legitimate) index, and the normalized value matched the stored value
so the row was skipped.
Strip any leading "data/" component before joining. This unblocks runtime
reads and lets _normalize_storage_paths rewrite the affected rows on next
startup — no manual migration needed.
Fixes "No such file or directory: '<data_dir>/data/profiles/...'" and
associated 404s on GET /audio/<id> after upgrading from 0.3.0 to 0.4.0.
Co-authored-by: Claude Opus 4.6 (1M context) <[email protected]>
- Show all voice profiles with unsupported ones grayed out (opacity) instead of hidden
- Clicking a grayed-out profile selects it and auto-switches the engine to a compatible one
- Sort supported profiles first, with info tip about compatibility at the bottom
- Scroll to selected profile after engine/sort changes with safe margin
- Fix engine desync on tab navigation by initializing form engine from store
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
When the model wasn't loaded, the app was closed mid-run, or a
generation otherwise errored out, the resulting "Failed" rows
accumulate in history and there was no way to remove them in bulk —
individual delete was the only option.
Adds a header row above the history list (only rendered when at
least one failed generation is present) with a "Clear failed" button
that opens a confirmation dialog, then calls a new
DELETE /history/failed endpoint which sweeps all status='failed'
rows (plus their version files / audio files on disk).
Closesjamiepine/voicebox#410
Co-authored-by: Claude Opus 4.6 <[email protected]>
Applies the compatibility-checker portion of #367. Adds a
check_cuda_compatibility() helper that compares the current device's
compute capability against torch.cuda._get_arch_list() and returns a
human-readable warning if the PyTorch build doesn't support it.
Wired into three places:
• HealthResponse gains a gpu_compatibility_warning field so clients
can surface the issue in the UI
• Startup logs the warning as WARN level
• _get_gpu_status() appends "[UNSUPPORTED - see logs]" to the GPU
label shown in settings
Skipped #367's other half — the switch from stable to nightly cu128
wheels across release.yml, build_binary.py, and justfile. That's
redundant with #401's TORCH_CUDA_ARCH_LIST=...12.0+PTX approach and
would introduce non-deterministic builds from shifting nightly
releases.
Co-Authored-By: nyzxor <[email protected]>
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Applies the cache_dir portion of #218. On Windows local setups, model
assets can split between .hf-cache/hub and .hf-cache/transformers when
Qwen3TTSModel.from_pretrained doesn't explicitly pin the cache root —
speech_tokenizer and preprocessor_config.json then fail to resolve
during load, causing 500s at generation time.
Routes both HF Hub and Transformers through hf_constants.HF_HUB_CACHE.
Skipped the torch_dtype= → dtype= rename from #218: transformers 4.36
(our minimum) doesn't accept the dtype alias, only 4.46+. Once we bump
the minimum we can make that change.
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Follow-up to #402. The sentinel is only removed inside the grace-period
"sentinel found" branch. When the HTTP /watchdog/disable request wins
the race (normal case on macOS/Linux, occasional on Windows), the
_watchdog_disabled=True check returns first and the sentinel is left on
disk indefinitely.
If a later session spawns a fresh server and the user exits without
"keep running", the new watchdog would find that stale sentinel during
its grace period and keep the server alive against user intent.
Wipe any pre-existing sentinel when the watchdog starts so only signals
written during this session's lifetime can influence grace-period
decisions.
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
The HTTP /watchdog/disable request races with process exit on Windows,
causing the watchdog to kill the server before the request arrives.
Added a .keep-running sentinel file as a reliable fallback:
- Tauri writes the file to data_dir before sending the HTTP request
- The watchdog checks for it during the grace period after detecting
parent death
- The file is removed after being read to avoid stale state
This approach works regardless of HTTP timing because file writes
complete synchronously before the Tauri process exits.
Fixes#372
Co-authored-by: Matt Van Horn <[email protected]>
* fix: add asyncio.Lock to prevent concurrent CUDA downloads
The startup auto-update task and the manual download endpoint can both
invoke download_cuda_binary() concurrently. Without mutual exclusion,
both coroutines write to the same temp file path, corrupting the
download. The progress-manager status check is a TOCTOU race because
the status is not set until after several synchronous checks complete.
Add a module-level asyncio.Lock acquired at the top of
download_cuda_binary() so only one download can proceed at a time.
* fix: fast-reject duplicate CUDA download when lock is held
Address CodeRabbit review feedback: check _download_lock.locked()
before awaiting the lock so concurrent callers return immediately
instead of queueing behind the first download. This prevents the
route handler from returning "started" to multiple callers when only
one download actually proceeds.
Two changes to address the race condition causing "Failed to split clip":
Backend (stories.py): Added with_for_update() to the item query in
split_story_item so concurrent requests for the same clip are
serialized via a row lock instead of racing.
Frontend (StoryTrackEditor.tsx): Guard handleSplit with
splitItem.isPending to prevent rapid double-clicks from firing
multiple mutations before the first completes.
Fixes#366
Co-authored-by: Matt Van Horn <[email protected]>
* fix(history): populate status/error/engine/model_size/is_favorited from DB
GET /history/{generation_id} was constructing HistoryResponse without
passing status, error, engine, model_size, or is_favorited from the
DB row. Since HistoryResponse.status defaults to "completed" in the
Pydantic model (models.py:141), this endpoint returned
status="completed" for every generation regardless of the actual DB
state — including jobs still in "loading_model" or "generating", and
even "failed" jobs.
This breaks any client polling /history/{id} for job completion:
the API lies about the status, so the only trustworthy success
signal becomes `audio_path` being non-empty. All other fields left
at their model defaults were similarly masked.
Fix: pass all fields through from the DB row, matching the pattern
used elsewhere in the codebase. The DB model (Generation in
database/models.py) already has all these columns.
* fix(history): apply NULL fallbacks to match list endpoint
Align the defensive mappings with services/history.py:206-223 so
both the single-item and list history endpoints handle legacy rows
with NULL status/engine/is_favorited identically. Without this,
HistoryResponse's non-Optional str/bool fields would raise a
pydantic ValidationError (500) on any row where these columns are
NULL — possible from direct SQL updates or past migrations.
Addresses review feedback on PR #394.
---------
Co-authored-by: malletfils <[email protected]>
Two small safety improvements:
1. Voice prompt cache (cache.py): add weights_only=True to torch.load()
so cached .prompt files are loaded using the safe unpickler instead of
the unrestricted pickle deserializer. This follows the PyTorch 2.6+
best practice of opting in to safe loading for all torch.load() calls.
2. SPA catch-all (app.py): replace str.startswith() path guard with
Path.is_relative_to(). The string prefix check passes for sibling
paths like /app/frontend_evil/ that share the /app/frontend prefix.
is_relative_to() correctly tests directory containment.
Follow-up to #361. The original fallback silently mapped unknown numpy
dtypes to torch.float32, which would reinterpret the memcpy'd bytes in
the wrong dtype and corrupt data (e.g. fp16 tensors from some TTS
engines) rather than erroring loudly.
- Hoist dtype_map out of the inner function so it's built once
- Add float16, complex64, complex128 mappings
- Raise TypeError on unknown dtype instead of silent float32 fallback
Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
torch is compiled against numpy 1.x. numpy 2.x changed the ABI version
returned by PyArray_GetNDArrayCVersion() (0x01000009 → 0x02000000), so
torch's is_numpy_available() always returns False and torch.from_numpy()
raises RuntimeError. This causes TTS generation to fail with:
ValueError: Unable to create tensor, you should probably activate
padding with 'padding=True'
Two fixes:
1. Pin numpy<2.0 in requirements.txt so new builds bundle a compatible
numpy version. (The existing comment already flagged this intention
but the upper bound was never added.)
2. Add a PyInstaller runtime hook (pyi_rth_numpy_compat.py) that installs
a ctypes memmove fallback for torch.from_numpy() at startup. Runtime
hooks run after FrozenImporter is registered so frozen torch is
importable. The fallback catches RuntimeError from the C-level ABI
check and copies the numpy array into a new tensor via raw memory copy,
bypassing the check entirely. This is a belt-and-suspenders fix that
works regardless of the bundled numpy version.
Co-authored-by: aimaaaimaa <[email protected]>
Co-authored-by: Claude Sonnet 4.6 <[email protected]>
Jamie Pine