Replace setup.py packaging with pyproject.toml, add uv

[ryanleary]

What does this PR do ?

Moves the project onto a standard PEP 621 build: one pyproject.toml replaces setup.py, setup.cfg, MANIFEST.in, reinstall.sh and requirements/. Adds uv for development environments, and removes the repository-root __init__.py.

The built wheel is unchanged apart from the first fix below — same py3-none-any tag, same version, same extras, same 13 runtime dependencies, every nemo_text_processing/ file preserved.

Problems this fixes

The wheel has been shipping the test suite. packages=find_packages() picks up tests/ because it has an __init__.py, and the exclude=['tools', 'tests', 'data'] argument was passed to setup(), where it is not a parameter and did nothing. Every release installs 247 test files into site-packages, including an importable top-level tests package. The new wheel drops them and is otherwise identical, file for file:

old: 2223 files      new: 1976 files
removed: 247         (all under tests/, none elsewhere)
added:   0
.tsv 798 -> 798      .far 5 -> 5      .fst 1 -> 1

[all] has been unsatisfiable on Python 3.8. The classifiers claimed 3.8 while the test extra pins black==25.1.0, which requires >=3.9. Nothing declared requires-python, so pip installed on 3.8 and failed later. Now declared >=3.9.

isort[requirements] requests an extra isort does not have, so the marker did nothing. Dropped.

What changes for users

./reinstall.sh becomes uv sync. pip install -e ".[all]" and pip install nemo_text_processing[all] are unaffected — this is a plain PEP 621 package, and uv is a convenience rather than a requirement. No uv.lock is committed.

Testing

tests/nemo_text_processing/en: 1613 passed, 2 failed — both pre-existing on an unmodified checkout (test_ordinal.py::test_norm_26_111st, test_serial.py::test_norm_32_your_otp_is_ab9453). black, isort and pytest settings resolve unchanged; uv build produces wheel and sdist with all data files and no MANIFEST.in.

---

Before your PR is "Ready for review"

Pre checks:

• Have you signed your commits? Use git commit -s to sign.
• Do all unittests finish successfully before sending PR? — 2 pre-existing unrelated failures. Sparrowhawk: N/A, no grammar is modified.
• If you are adding a new feature: Have you added test cases — N/A, no new feature.
• Have you added __init__.py for every folder and subfolder — no folders added; one stray root-level __init__.py removed.
• Have you followed codeQL results and removed unused variables and imports?
• Have you added the correct license header to all newly added Python files — pyproject.toml carries it.
• Remove import guards (try import: ... except: ...) if not already done — none introduced.
• If you added a new language or a new feature please update the NeMo documentation — N/A.
• Have you added your language support to pynini_export.py — N/A.

PR Type:

• New Feature
• Bugfix
• Documentation
• Test

🤖 Generated with Claude Code