Reviewer feedback: remove AudioContext destination echo, escape model names, add RMS gate to skip silence, close mkstemp fd, clean up mic on WS close. Tests updated for the RMS gate.
Full HTML dashboard (dark theme, vanilla JS, no external deps) served at / and /dashboard with five tabs: system status (admin), file upload-to-transcribe with progress/result/downloads, job history with cancel/result modal, realtime mic demo over the existing WebSocket, and API key create/list.
Backend: POST/GET /v1/keys (admin; raw key returned once, digest-only storage); KeyStore.list_keys(); EngineOwner.emit_hypothesis is now a real implementation (PCM16 chunk -> WAV -> realtime-lane decode with the single GPU lock) instead of a stub.
Notebook: tunnel cell links /dashboard and smoke-checks the HTML.
+ 6 tests (dashboard public HTML, key create/list/scope, WAV header, emit_hypothesis cleanup). 153 tests pass, ruff clean, JS syntax verified with node --check.
The bench and CLI kept emitting BLM for vLLM. Three layers:
1. rules.py: add default rule BLM -> vLLM (word-boundary, case-insensitive) so the default post_mode=rules path restores it everywhere.
2. benchmark/runner.py: the bench never ran postprocessing - it measured raw engine text, so entity retention (66.7%) never reflected rules/glossary. _transcribe_clip now builds Segments and runs run_postprocess(settings, glossary) before metrics; manifest top-level glossary {pattern: replacement} is supported and recorded in run_config (post_mode/glossary).
3. glossary plumbing: CLI transcribe gains --glossary KEY=VALUE (repeatable, parsed + validated); BatchPipeline.run accepts glossary= and passes it to run_postprocess; the in-proc worker forwards job.options.glossary or post_correction.
Notebook: transcribe cells pass --glossary BLM=vLLM, bench manifest carries glossary, postprocess section documents rules/glossary/hotword.
+ 9 unit tests (BLM rule, boundary/case behavior, bench postprocess + glossary entity retention). 147 tests pass, ruff clean.
Reviewer feedback: cut worst-case stall from ~200s to ~60s (urllib 2x, DoH 4x with max-time 10); when curl reports 000 (connect failure) prefer stderr for the diagnostic; print an explicit note when falling back to DoH after a resolved-DNS urllib failure.
Colab VM DNS fails to resolve fresh trycloudflare hostnames (Name or service not known) while the tunnel itself works from any real browser. Verification now: 1) diagnoses VM DNS via socket.gethostbyname, 2) uses urllib when DNS resolves, 3) falls back to curl --doh-url (Cloudflare DNS over HTTPS) to bypass the VM resolver, 4) otherwise prints the docs URL with an explicit note that a minted URL means the tunnel is already connected. Troubleshooting section documents the DNS limitation.
Reviewer feedback: wrap the install-cell import verification in try/except so a walk-up failure cannot crash the cell (previously always-succeeding), and keep the last exception detail in the tunnel external-verification failure message.
Colab run 5 failed at cell 12 (import luke_scribe.config) because hatchling editable installs write a .pth pointing at src/ that the interpreter only reads at startup; a long-running Colab kernel started before pip install -e never sees it. Subprocesses (CLI) work, but kernel-side imports fail.
Fix: install cell now walks up to the repo root and inserts src/ into the kernel sys.path, verifying with import luke_scribe.config. Cell 12 has a defensive guard of the same kind. Tunnel external verification now retries up to 6x2s for DNS propagation. Troubleshooting section documents the issue.
Reviewer feedback: guard the 120s URL poll with shutil.which(cloudflared) so the failure case returns immediately; merge import re into the cell top import; note Cloudflare browser-check interstitials can fail urllib verification even when the link works in a real browser.
Server cell sets LUKESCRIBE_TUNNEL=cloudflare so the app built-in CloudflareTunnel (trycloudflare quick tunnel) starts alongside the server. After health+auth checks the cell polls server.log for the trycloudflare URL, prints it with /docs dashboard and /health links, and verifies external reachability.
cloudflared binary is downloaded in the install cell (graceful skip on failure). Troubleshooting section documents the temporary/public nature of the link.
Bench clip failed in Colab because entity_retention calls ent.get() but
the manifest used plain strings -> AttributeError -> clip counted as
failure (failure_rate 1.0). Notebook now builds entities as
{canonical, surface, start_char, end_char} from the reference text and
writes a reference file for meaningful metrics.
Colab run 4 (A100): same namedtuple bug as segments — faster-whisper
returns TranscriptionInfo (namedtuple) but batch.py reads
outcome['info'].get('language') -> AttributeError 'TranscriptionInfo'
object has no attribute 'get' on every real transcription, failing the
CLI, API auto-worker job, worker drain, and bench clip alike.
FasterWhisperEngine now normalizes info to a dict at the boundary
(_to_dict_info: _asdict -> dataclasses.asdict -> known-field fallback).
+ 2 unit tests (namedtuple/dict info); 138 tests pass, ruff clean.
Colab run 3 (A100) surfaced three GPU/API-path bugs mocks couldn't catch:
1. faster-whisper yields namedtuple Segments, but batch/bench consume
them as dicts (.get) -> AttributeError 'Segment' has no attribute
'get' on every real transcription. Engine now normalizes segments
to dicts (_to_dict_segments) at the boundary.
2. API TranscribeOptions carries engine-irrelevant keys (formats,
timestamps, glossary_id, post_correction, diarize); worker's
TranscriptionOptions(**job.options) crashed with TypeError. Worker
now filters job.options to TranscriptionOptions.__slots__.
3. in-proc server never consumed its own queue (jobs stayed queued
forever). Added opt-in Settings.auto_worker (default off): lifespan
starts a daemon Worker thread for inproc backend, stopped on
shutdown. Notebook enables it via LUKESCRIBE_AUTO_WORKER=true so the
API upload -> completed flow works end to end.
Notebook: bench manifest now uses clips schema (audio_path/duration_sec/
entities); cell 22 reads error_message/error_code; upload poll window
raised to 4min (first-run model download).
+ 5 tests (namedtuple/dict segments, API-style options, auto_worker
on/off); 136 tests pass, ruff clean.
Colab A100 run exposed a GPU-only bug: DeviceManager returns
selected_device='cuda:0' and the engine passed it verbatim to
faster-whisper, but CTranslate2 only accepts device='cuda' with a
separate device_index arg -> 'unsupported device cuda:0' on every GPU
transcription. Fix: FasterWhisperEngine._split_device() splits
'cuda:N' -> ('cuda', N) and passes device_index to WhisperModel.
Notebook: server cell now pkills stale servers (old process holds
port 8000 and 401s on new keys since KeyStore loads at startup),
sets LUKESCRIBE_API_KEY_FILE explicitly, and verifies auth with
RAW_KEY before proceeding; worker cell guards read_result() is None.
+ 4 unit tests (device split contract), 131 tests pass, ruff clean.
nvidia-cudnn-cu12/cublas-cu12 are namespace packages with no
__file__, so os.path.dirname() raised TypeError and the install cell
aborted. Locate libcublas.so/libcudnn.so under site-packages with
find and build LD_LIBRARY_PATH from those. Also add a CTranslate2
GPU check cell (ctranslate2.get_cuda_device_count) right after
install so GPU usability is confirmed before transcription.
Colab now ships CUDA 13.0 (driver 580); CTranslate2 wheels are CUDA 12
so GPU init fails with 'unsupported device cuda:0'. Install
nvidia-cublas-cu12/nvidia-cudnn-cu12 and set LD_LIBRARY_PATH in the
install cell; transcription cell reports failures cleanly.
API smoke: api_keys.json stores only digests (raw key shown once), so
the notebook now captures RAW_KEY at creation and uses it for upload
instead of reading the digest file (fixes 401).
Cell 1 now detects an existing /content/luke_scribe/.git and runs
git fetch + checkout feat/full-platform + pull --ff-only instead of
rm -rf + clone, so re-running the notebook updates instead of wiping
local changes.
Colab 'python3 -m venv' fails with ensurepip error (no .venv created,
so every subsequent cell hit 'command not found'). Switch to system pip
(Colab standard) and run the API server via nohup background with log
fallback diagnostics.
CPU-only dev env verified via mocks; the notebook runs the full real
pipeline on Colab Pro T4: clone → ffmpeg/venv install → detect (GPU
capability tier) → 127 unit/integration tests → sample TTS (KO+EN
tech terms) → real faster-whisper transcription → hotword/postprocess
→ API smoke → benchmark.
[gstack-context]
Decisions: Validate the transcription core and benchmark gate before API, queue, and realtime work; preserve five expansion contracts.
Remaining: Build the benchmark dataset and implement detect/transcribe/bench.
Skill: /office-hours
[/gstack-context]
Populate the previously-empty .omc/project-memory.json so teammates and
future OMC sessions inherit context: 4 user directives (SoT location,
greenfield/next-step, locked design decisions, measurement-gated residual),
3 notes (architecture, tech stack, env), and the decided tech stack.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fold post-plan decisions into the spec and consensus plan:
- Q1 deploy HW: undecided/mixed → delegate to hardware-adaptive auto-sizing
- Q2 model strategy: collapse to single turbo model if P1 bench entity ≥95%
- Q3 cancellation: cooperative (segment-boundary) is sufficient; no hard-kill
- Q4 concurrency N: delegate to boot-time auto-sizing (AC-8 = ≤5s within auto N)
Recompute clarity with the deep-interview model (Goal 0.96 / Constraint 0.95
/ Success 0.95 → Total 0.954): ambiguity ~10% → ~5%. Residual is now entirely
measurement/code-gated (AC-4 R-WER baseline, hybrid→single confirmation,
CT2 GIL) — next lever is P1 bench, not further interview.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
lukehemmin
