Three phases that only make sense together: a job that names a worker,
a worker that reports honestly what it can actually run, and the small
defects that made both lie.
**Pinning** (Phase 1). `pinned_worker_id` is now honoured in both places
that choose a worker — `eligible_workers` and `select_worker` build
independent lists, so applying it to one silently leaked work onto
whichever machine was least busy. The pin persists across a restart via
an additive column, deliberately not alembic (justified in the code, per
the precedent already in db.py): quitting mid-render used to drop it
without a word. `max_attempts=1` was rejected as the mechanism — it makes
the FIRST failure terminal, including the penalty-free ones a stale
advisory view produces routinely.
Cancel now actually reaches the worker. `WorkerServicer.cancel` had zero
callers, so cancelling released the slot while the GPU thread kept
running, and a late result could resurrect the task as COMPLETED —
`commit_result` assigned that state directly, bypassing the transition
table where CANCELLED is terminal by construction.
**Honest capabilities** (Phase 4). A worker now probes whether weights
are actually present, and a job stops BEFORE dispatch with a typed 409
naming the model and the machine, instead of failing mid-task. The probe
fails OPEN: `is_cached`/`cache_is_complete` cannot see a user-managed
clone outside the HF layout, so only a positive "absent" refuses.
Refusing an engine that works today would break the compatibility
promise. `pool.supports` deliberately still ignores `downloaded` — had it
not, the scheduler would drop the worker and answer with a terminal
NO_CAPABLE_WORKER, which tells the user to check their install when the
truth is one download away. The frontend no longer offers "Report this
bug" for that state; it offers the download.
Catalog tags resolve against the TARGET's OS/arch/backend, not this
machine's. From a Mac control plane, a CUDA worker's model list was
showing the mlx-community repos it cannot run and hiding the ones it
needs.
**And the quiet ones** (Phase 0 leftovers): a model's human label rides
its own proto field so renaming it cannot orphan breaker history; an
empty model_id no longer forks the capacity slot key into two slots for
one model; the idle sweep cannot evict an engine out from under a live
LOCAL render.
Verified on real hardware, which is the only verification that has ever
caught anything here: 2025 characters, default settings, routed to an
RTX 4090 over the wire — 100% GPU utilisation on the remote box, 119.6 s
of 24 kHz audio returned in 16.6 s, 5.7 MB delivered out of band through
the artifact path rather than the control stream.
Backend 5259 passed, frontend 1808 passed.
Only the TTS model (~2.4 GB) is required on first run; ASR models are
per-platform curated picks (curated_on in models.yaml) installed on demand.
Every transcription surface returns a typed asr_model_missing error with a
one-click download CTA instead of silently pulling multi-GB Whisper weights.
Settings -> Models is a grouped, platform-aware catalog. New guided
permissions UX (wizard System Check + Settings -> Permissions + mic
pre-flight) with native mic-state checks and OS settings deep-links. New
parakeet-mlx engine brings Parakeet TDT v3 to Apple Silicon (language-gated
capture preference so multilingual dictation never regresses). Docs:
expressive-speech page, Flush/Unload + CPU-fallback triage, clone-length FAQ.
Hardening: preflight fails open for custom model pins, ROCm curation no
longer inherits NVIDIA picks, Windows mic probe reads the NonPackaged
consent key, CaptureWidget setup race fixed, offline-cache CI simulation
fixes so empty-cache runners stay green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Palash Debnath