Files
karaoke-gen/backend
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
..

Backend Development Setup

One-Time Setup

1. Create Virtual Environment

cd /Users/andrew/Projects/karaoke-gen
python3 -m venv backend/venv

2. Activate Virtual Environment

source backend/venv/bin/activate

3. Install Dependencies

IMPORTANT: Use Python 3.12 (not 3.13/3.14 - pydantic has compatibility issues)

# Backend dependencies
pip install -r backend/requirements.txt

# Install karaoke_gen in development mode
pip install -e .

# Install lyrics_transcriber submodule
pip install -e lyrics_transcriber_local/

# Install Google API client (for YouTube upload)
pip install google-api-python-client google-auth-httplib2 google-auth-oauthlib

4. Add Shell Alias (Optional)

Add to ~/.zshrc:

alias backend='cd /Users/andrew/Projects/karaoke-gen && source backend/venv/bin/activate'

Then you can just run:

backend
# Now you're in the project directory with venv activated!

Daily Workflow

Before Making Changes

# Activate venv
source backend/venv/bin/activate

# Or if you set up the alias:
backend

After Making Changes

# 1. Run full validation (catches import errors!)
python3 backend/validate.py

# 2. If validation passes, deploy
./scripts/deploy.sh

Validation Options

source backend/venv/bin/activate
python3 backend/validate.py

Catches:

  • ✅ Syntax errors
  • ✅ Import errors (like the processing_service issue!)
  • ✅ Configuration issues
  • ✅ FastAPI app creation problems
  • ✅ Missing dependencies

This would have caught the ModuleNotFoundError!

Quick Check (No venv needed)

./backend/quick-check.sh

Catches:

  • ✅ Syntax errors
  • ✅ Missing files
  • ✅ Some obvious import issues

Good for quick checks, but won't catch all import errors


Example: Complete Development Session

# 1. Activate environment
backend  # or: source backend/venv/bin/activate

# 2. Make your changes
vim backend/api/routes/jobs.py

# 3. Validate (catches import errors!)
python3 backend/validate.py
# ✅ All validations passed!

# 4. Deploy
./scripts/deploy.sh

# 5. Test
export BACKEND_URL="https://karaoke-backend-ipzqd2k4yq-uc.a.run.app"
export AUTH_TOKEN=$(gcloud auth print-identity-token)
curl -H "Authorization: Bearer $AUTH_TOKEN" $BACKEND_URL/api/health

What About karaoke_gen Imports?

The backend imports from karaoke_gen (the CLI package):

from karaoke_gen.karaoke_gen import KaraokePrep
from karaoke_gen.audio_processor import AudioProcessor

Options:

source backend/venv/bin/activate
pip install -e .

This makes karaoke_gen available in the venv without copying files.

Option 2: Use system Python

Your system Python already has karaoke_gen installed, so validation will work even if the venv doesn't have it (Python will fall back to system packages).


Running the Backend Locally

# Activate venv
source backend/venv/bin/activate

# Set required environment variables
export GOOGLE_CLOUD_PROJECT="nomadkaraoke"
export GCS_BUCKET_NAME="karaoke-gen-storage-nomadkaraoke"

# Optional: Point to local credentials
export GOOGLE_APPLICATION_CREDENTIALS="/path/to/service-account-key.json"

# Run the server
cd backend
uvicorn main:app --reload --port 8080

Visit: http://localhost:8080/api/health


Troubleshooting

"No module named 'fastapi'"

Fix:

source backend/venv/bin/activate
pip install -r backend/requirements.txt

"No module named 'karaoke_gen'"

Fix:

source backend/venv/bin/activate
cd /Users/andrew/Projects/karaoke-gen
pip install -e .

Validation passes locally but fails in Cloud Run

Check:

  1. Is requirements.txt up to date?
  2. Did you add a new dependency without adding it to requirements.txt?
  3. Run: pip freeze | grep <package-name>

Updating Dependencies

When you add new packages:

# 1. Install in venv
source backend/venv/bin/activate
pip install new-package

# 2. Update requirements.txt
pip freeze > backend/requirements.txt

# 3. Or manually add just the new package:
echo "new-package==1.2.3" >> backend/requirements.txt

# 4. Validate
python3 backend/validate.py

# 5. Deploy
./scripts/deploy.sh

Summary

Best practice workflow:

  1. One-time setup: Create venv and install dependencies
  2. Daily: Activate venv when working on backend
  3. Before deploy: Run python3 backend/validate.py (catches import errors!)
  4. Deploy: Use ./scripts/deploy.sh

The import error would have been caught if you ran validation in a venv! ✅