JMR-devandClaude Opus 5
2cfd7ad087
Artifacts on disk are the source of truth. state.json deliberately has no top-level "stage" field -- persisting one is how "marked done but the file is gone" bugs happen -- so the resume point is computed from what verifies. The distinction that makes --no-retain safe is deleted_by_policy vs missing. verify_* short-circuits on a policy deletion before touching the filesystem, because probing a deliberately absent file would raise and degrade the whole feature into "re-download everything". A .part without its .aria2 control file is treated as unresumable: aria2 writes segments out of order, so such a file is sparse with holes rather than a valid prefix, and resuming from its length yields a corrupt video. Planning walks stages backwards. A policy deletion satisfies a stage that is not re-running, but not one that is -- so --force-stage transcribe correctly walks back to re-download. Saved segments let a deleted subtitle file be re-rendered without re-transcribing a long recording. state.json is written tmp -> fsync -> replace -> fsync(dir), with a test that a failed replace leaves the previous record intact. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
98 lines
2.4 KiB
TOML
98 lines
2.4 KiB
TOML
[build-system]
|
|
requires = ["hatchling"]
|
|
build-backend = "hatchling.build"
|
|
|
|
[project]
|
|
name = "ccn-transcribe"
|
|
version = "0.1.0"
|
|
description = "Download videos, extract FLAC, and transcribe with Whisper on whatever GPU is present."
|
|
requires-python = ">=3.12"
|
|
dependencies = [
|
|
"click>=8.1,<9",
|
|
"numpy>=2.5,<3",
|
|
"huggingface-hub>=1.31,<2",
|
|
"yt-dlp[default,deno]>=2026.8.19",
|
|
"yt-dlp-ejs>=0.8",
|
|
"openvino==2026.3.1",
|
|
"openvino-genai==2026.3.1.0",
|
|
]
|
|
|
|
[project.scripts]
|
|
ccn-transcribe = "ccn_transcribe.cli:main"
|
|
|
|
[dependency-groups]
|
|
dev = [
|
|
"pytest>=8",
|
|
"pytest-cov>=5",
|
|
"ruff>=0.6",
|
|
"pyright>=1.1.380",
|
|
"mypy>=1.11",
|
|
"pylint>=3.2",
|
|
"pre-commit>=3.8",
|
|
]
|
|
|
|
[tool.hatch.build.targets.wheel]
|
|
packages = ["src/ccn_transcribe"]
|
|
|
|
[tool.ruff]
|
|
line-length = 100
|
|
src = ["src", "tests"]
|
|
# ruff format also restyles Python blocks inside Markdown; docs are not ours to restyle.
|
|
extend-exclude = ["*.md"]
|
|
|
|
[tool.ruff.lint]
|
|
select = ["E", "F", "W", "I", "N", "UP", "B", "A", "C4", "SIM", "PTH", "RET", "ARG", "TID", "TC", "RUF", "S"]
|
|
# S607: binaries are looked up on PATH on purpose and preflight-checked with
|
|
# shutil.which; hardcoding absolute paths would be less portable, not safer.
|
|
ignore = ["S607"]
|
|
|
|
[tool.ruff.lint.per-file-ignores]
|
|
# S101: asserts are the point of a test file.
|
|
"tests/*" = ["ARG001", "ARG002", "S101"]
|
|
# Throwaway benchmarks, not pipeline code.
|
|
"scripts/*" = ["S603"]
|
|
|
|
[tool.pyright]
|
|
include = ["src", "tests"]
|
|
typeCheckingMode = "strict"
|
|
venvPath = "."
|
|
venv = ".venv"
|
|
|
|
[tool.mypy]
|
|
strict = true
|
|
files = ["src", "tests"]
|
|
|
|
[tool.pylint.design]
|
|
max-attributes = 15
|
|
|
|
[tool.pylint.main]
|
|
# W0621: pytest fixtures shadow their names by design.
|
|
# C1803: asserting == () is more precise than truthiness about what a call returns.
|
|
# R0903: dataclasses and pytest grouping classes legitimately have few methods.
|
|
disable = [
|
|
"C0114", "C0115", "C0116", "C0301",
|
|
"W0611", "W0612", "W0621",
|
|
"R0801", "R0903", "C1803",
|
|
]
|
|
|
|
[tool.pytest.ini_options]
|
|
testpaths = ["tests"]
|
|
# NOTE: --cov-fail-under deliberately lives in the pre-commit hook, not here, so a
|
|
# single-file TDD run (`uv run pytest tests/test_x.py`) is not blocked by coverage.
|
|
addopts = "-q"
|
|
|
|
[tool.coverage.run]
|
|
branch = true
|
|
source = ["src/ccn_transcribe"]
|
|
|
|
[tool.coverage.report]
|
|
fail_under = 100
|
|
show_missing = true
|
|
exclude_also = [
|
|
"if TYPE_CHECKING:",
|
|
'if __name__ == "__main__":',
|
|
"@overload",
|
|
"raise NotImplementedError",
|
|
"class .*\\(Protocol\\):",
|
|
]
|