Files
karaoke-gen/pyproject.toml
T
Andrew Beveridge ☄️andClaude Opus 5.5 acb02e138a fix: un-pin audio-separator to 0.47.1 with RoFormer overlap=2 (v0.278.16) (#1196)
@coderabbitai ignore

## Why
BACKLOG "audio-separator 0.45–0.47 regression: mel-band RoFormer karaoke
models 3–5× slower on L4". The root cause is **python-audio-separator
#299 (0.47.0)**, not #298. #299 made RoFormer `overlap` a *window count*
that defaults to each model YAML's `num_overlap`. Before that, `overlap`
was effectively an 8-second step, which meant almost no overlap. The
karaoke models' YAMLs set `num_overlap` 4/8, so a 3-minute track went
from 23 forward passes to 87 (aufr33) and 124 (gabox_v2, becruily). That
matches the measured 25→75s, 26→149s and 27→157s. 0.45.0 and 0.46.0
still had the old step formula.

python-audio-separator #313 (released as **0.47.1**) keeps #299. It adds
a per-model forward-pass budget test on every PR and a GPU per-pass
timing test (T4: 1.21 s/pass), plus a README note. Andrew chose
overlap=2 for gen: smoother chunk seams than 0.44's no-overlap, at about
1.5× the 0.44 compute instead of 4–6×.

## What
- `karaoke_gen/audio_processor.py`: `SEPARATOR_MDXC_PARAMS` (overlap=2)
goes to all four `Separator(...)` constructions (stage 1, stage 2
karaoke ensemble, single-model paths) **only when audio-separator ≥
0.47**. On older versions `overlap=2` meant a 2-second step, which is 4×
slower, so they keep their defaults. Ensembles load each model through
`load_model`, so the param reaches every model in the preset.
- `backend/Dockerfile.gpu-base`: `AUDIO_SEPARATOR_VERSION=0.47.1`, with
the comment rewritten to explain the root cause. 0.47.1 needs
`torch>=2.3` on Linux, so the pre-installed `torch==2.6.0+cu126`
(required by the L4 driver) is kept.
- The poetry lock stays on 0.44.1. The version-aware params keep CI and
CPU images on the old semantics, and there's no torch churn.

## Expected cost per ~3-min track (forward passes)
| model | 0.44.2 | 0.47 YAML default | 0.47 + overlap=2 |
|---|---|---|---|
| karaoke_aufr33 | 23 | 87 | 44 |
| karaoke_gabox_v2 | 23 | 124 | 32 |
| karaoke_becruily | 23 | 124 | 32 |
| instrumental fv7z / resurrection | 23 | 32 / 21 | 32 / 21 |

## Verification plan (before /backlog-finish)
After the deploy, time the separation of the daily E2E or a real job on
the audio-separation-job logs. I expect roughly 1.5× the 0.44.2
karaoke-stage times (~40–50s per karaoke model rather than 75–157s). If
it's worse, revert the pin to 0.44.2. That's one line.

## Testing
- `tests/unit/test_audio_processor_local_gpu.py`: version gating
(0.47.x/0.48/1.0 → overlap 2; 0.44/0.46 → defaults; unknown → defaults;
fresh copy), and both stages' constructors receive `mdxc_params.overlap
== 2`.
- `make test`: the backend unit suite passes. The only failures are 8
`test_emulator_integration.py` tests, caused by my local environment:
another session's emulator setup is on :8080/:4443, and they fail
identically on untouched main. CI is green on main.
- Local CodeRabbit: 0 findings.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

https://claude.ai/code/session_01Jw21syo4HizJJjYw2mg2Vp

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-19 15:09:41 -04:00

205 lines
6.9 KiB
TOML

[tool.poetry]
name = "karaoke-gen"
version = "0.278.16"
description = "Generate karaoke videos with synchronized lyrics. Handles the entire process from downloading audio and lyrics to creating the final video with title screens."
authors = ["Andrew Beveridge <andrew@beveridge.uk>"]
license = "MIT"
readme = "README.md"
packages = [
{ include = "karaoke_gen" },
{ include = "backend" }
]
include = [
# The Next.js static export is generated at build time and gitignored (see #876).
# Poetry's plain-string include respects .gitignore and would DROP a gitignored
# path from the wheel; the dict form with an explicit `format` overrides .gitignore
# so the built frontend is still bundled (needed for the CLI local-review UI).
# Wheel-only: the sdist would balloon past PyPI's 10 GB cap, so bundle in the wheel only.
{ path = "karaoke_gen/nextjs_frontend/out/**/*", format = ["wheel"] },
"karaoke_gen/lyrics_transcriber/output/fonts/*",
"karaoke_gen/lyrics_transcriber/output/cdgmaker/images/*",
"karaoke_gen/lyrics_transcriber/output/cdgmaker/transitions/*",
# Skip/include lists for the YouTube description backfill worker (must ship in
# the Cloud Run image so the scheduled worker can read them).
"backend/data/youtube_backfill/*.txt"
]
homepage = "https://github.com/nomadkaraoke/karaoke-gen"
repository = "https://github.com/nomadkaraoke/karaoke-gen"
documentation = "https://github.com/nomadkaraoke/karaoke-gen/blob/main/README.md"
[tool.poetry.dependencies]
python = ">=3.10,<3.14"
torch = ">=2.7"
requests = ">=2"
beautifulsoup4 = ">=4"
flacfetch = ">=0.9.0"
lyricsgenius = ">=3"
fetch-lyrics-from-genius = ">=0.1"
pillow = ">=10.1"
pyinstaller = ">=6.3"
google-api-python-client = "2.193.0"
google-auth = "2.49.1"
google-auth-oauthlib = "1.3.0"
google-auth-httplib2 = "0.3.0"
thefuzz = ">=0.22"
numpy = ">=2"
audio-separator = { version = ">=0.43.0", extras = ["cpu"] }
lyrics-converter = ">=0.2.1"
# lyrics_transcriber is now a subpackage of karaoke_gen (karaoke_gen/lyrics_transcriber/)
# Dependencies from vendored lyrics_transcriber:
mutagen = ">=1.47" # For LRCLIB provider to get track duration for better matching
python-slugify = ">=8"
syrics = ">=0"
karaoke-lyrics-processor = ">=0.6"
dropbox = ">=12"
spacy = ">=3.8.7"
srsly = ">=2.5.1"
tqdm = ">=4.67"
python-levenshtein = ">=0.26"
transformers = ">=4.47"
metaphone = ">=0.6"
nltk = ">=3.9"
spacy-syllables = ">=3"
syllables = ">=1"
fonttools = ">=4.55"
ollama = ">=0.4.7"
shortuuid = ">=1.0.13"
nest-asyncio = ">=1.5"
openai = ">=1.63.2"
langgraph = ">=0.2.0"
langchain = ">=0.3.0"
langchain-core = ">=0.3.0"
langchain-openai = ">=0.2.0"
langchain-anthropic = ">=0.2.0"
anthropic = ">=0.109.0" # direct dep: auto-correct structured outputs (output_config)
langchain-ollama = ">=0.2.0"
langchain-google-genai = ">=2.0.0"
langfuse = ">=3.0.0"
jiwer = ">=3.0.0"
kbputils = "^0.0.16"
attrs = ">=24.2.0"
cattrs = ">=24.1.2"
toml = ">=0.10"
argparse = ">=1.4.0"
psutil = "^7.0.0"
pyperclip = "*"
pytest-asyncio = "*"
ffmpeg-python = "^0.2.0"
python-multipart = "^0.0.20"
python-docx = ">=1.1.0"
yt-dlp = ">=2024.0.0"
matplotlib = ">=3"
# Backend API dependencies
fastapi = ">=0.104.0"
uvicorn = { version = ">=0.24.0", extras = ["standard"] }
pydantic = ">=2.5.0"
pydantic-settings = ">=2.1.0"
httpx = ">=0.25.0"
tenacity = "^8.2.0" # Retry logic for external service calls
google-cloud-firestore = "2.26.0"
google-cloud-compute = "^1.0.0"
google-cloud-storage = "3.10.1"
google-cloud-secret-manager = "2.26.0"
google-cloud-monitoring = "2.29.1" # Cloud Monitoring for YouTube API quota metrics
google-cloud-tasks = "2.21.0" # Cloud Tasks for worker coordination (Phase 1 scalability)
google-cloud-run = "0.15.0" # Cloud Run Jobs for long-running video encoding (Phase 2 scalability)
google-cloud-logging = ">=3.9.0" # Cloud Logging for error monitoring
python-dotenv = ">=1.0.0"
pydub = ">=0.25.1"
# Google / Apple ID-token verification (JWKS) for mobile sign-in
pyjwt = { version = ">=2.8.0", extras = ["crypto"] }
# Payment and email services
stripe = "^14.1.0"
email-validator = ">=2.0.0" # For Pydantic EmailStr validation
# Web Push notifications
pywebpush = ">=2.0.0"
# OpenTelemetry for distributed tracing
opentelemetry-api = ">=1.20.0"
opentelemetry-sdk = ">=1.20.0"
opentelemetry-instrumentation-fastapi = ">=0.41b0"
opentelemetry-instrumentation-httpx = ">=0.41b0"
opentelemetry-instrumentation-logging = ">=0.41b0"
opentelemetry-exporter-gcp-trace = ">=1.6.0"
opentelemetry-resourcedetector-gcp = ">=1.6.0a0"
# Optional: Local Whisper transcription (install with: pip install karaoke-gen[local-whisper])
whisper-timestamped = { version = ">=1.15.0", optional = true }
aiohttp = "^3.13.2"
[tool.poetry.group.dev.dependencies]
black = ">=23"
poetry = "*"
pytest = ">=7.0"
pytest-cov = ">=4.0"
pytest-mock = ">=3.10"
pytest-asyncio = "*"
# Make dev dependencies available as optional extras for pip install
# These need to be declared as optional dependencies first
[tool.poetry.dependencies.black]
version = ">=23"
optional = true
[tool.poetry.dependencies.pytest]
version = ">=7.0"
optional = true
[tool.poetry.dependencies.pytest-cov]
version = ">=4.0"
optional = true
[tool.poetry.dependencies.pytest-mock]
version = ">=3.10"
optional = true
[tool.poetry.extras]
# Local Whisper transcription support - optional, heavy dependencies
# Install with: pip install karaoke-gen[local-whisper]
# For CPU-only installation (recommended for systems without GPU):
# pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu
# pip install karaoke-gen[local-whisper]
local-whisper = ["whisper-timestamped"]
# Development/testing dependencies (pytest-asyncio is already a main dependency)
# Install with: pip install -e ".[dev]" or pip install -e ".[local-whisper,dev]"
dev = ["black", "pytest", "pytest-cov", "pytest-mock"]
[tool.poetry.scripts]
karaoke-gen = 'karaoke_gen.utils.gen_cli:main'
karaoke-gen-remote = 'karaoke_gen.utils.remote_cli:main'
karaoke-bulk = 'karaoke_gen.utils.bulk_cli:main'
[tool.black]
line-length = 140
[build-system]
requires = ["poetry-core"]
build-backend = "poetry.core.masonry.api"
[tool.pytest.ini_options]
testpaths = ["tests/unit", "tests/integration", "backend/tests"]
python_files = ["test_*.py"]
addopts = [
"--asyncio-mode=auto",
"-v",
"--tb=short",
]
markers = [
"slow: marks tests as slow (deselect with '-m \"not slow\"')",
"integration: marks tests as integration tests",
"asyncio: mark test as async",
"emulator: marks tests requiring GCP emulators",
]
filterwarnings = [
"ignore::DeprecationWarning",
"ignore:'audioop' is deprecated:DeprecationWarning",
"ignore::RuntimeWarning",
"ignore::UserWarning",
]
[tool.coverage.run]
# Exclude lyrics_transcriber subpackage from coverage since it was a separate vendored package
# with its own test suite - including it in karaoke_gen coverage would unfairly lower the percentage
omit = [
"karaoke_gen/lyrics_transcriber/*",
]