Fix installation instructions for Python 3.13 incompatibility and improve the sanity check example

[kevin85421]

Motivation & Modifications

Modification 1: Change from Python 3.13 to 3.12.

On Python 3.13, uv pip install -v -e . fails because outlines-core v0.1.26 is incompatible with it.

help: `outlines-core` (v0.1.26) was included because `sglang-omni` (v0.1.0) depends on `sglang` (v0.5.8) which depends on `outlines` (v0.1.11) which depends on `outlines-core`

• outlines-core only has wheels for Python 3.9-3.12.

  ➜  ~ curl -s https://pypi.org/pypi/outlines-core/0.1.26/json \
    | jq -r '.urls[].filename' | sort
  outlines_core-0.1.26-cp310-cp310-macosx_10_12_x86_64.whl
  outlines_core-0.1.26-cp310-cp310-macosx_11_0_arm64.whl
  outlines_core-0.1.26-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
  outlines_core-0.1.26-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
  outlines_core-0.1.26-cp310-cp310-win_amd64.whl
  outlines_core-0.1.26-cp310-cp310-win32.whl
  outlines_core-0.1.26-cp311-cp311-macosx_10_12_x86_64.whl
  outlines_core-0.1.26-cp311-cp311-macosx_11_0_arm64.whl
  outlines_core-0.1.26-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
  outlines_core-0.1.26-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
  outlines_core-0.1.26-cp311-cp311-win_amd64.whl
  outlines_core-0.1.26-cp311-cp311-win32.whl
  outlines_core-0.1.26-cp312-cp312-macosx_10_13_x86_64.whl
  outlines_core-0.1.26-cp312-cp312-macosx_11_0_arm64.whl
  outlines_core-0.1.26-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
  outlines_core-0.1.26-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
  outlines_core-0.1.26-cp312-cp312-win_amd64.whl
  outlines_core-0.1.26-cp312-cp312-win32.whl
  outlines_core-0.1.26-cp39-cp39-macosx_10_12_x86_64.whl
  outlines_core-0.1.26-cp39-cp39-macosx_11_0_arm64.whl
  outlines_core-0.1.26-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
  outlines_core-0.1.26-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
  outlines_core-0.1.26-cp39-cp39-win_amd64.whl
  outlines_core-0.1.26-cp39-cp39-win32.whl
  outlines_core-0.1.26.tar.gz
  

The next step is to bump outlines in upstream SGLang and then update SGLang Omni to pin to the new SGLang release.

Modification 2: Change max_tokens from 16 to 1024

The example uses "max_tokens": 16, which is too low for the model to finish answering, so it does not serve as a meaningful sanity check.

Checklist

• Format your code according with pre-commit.
• Add unit tests.
• Update documentation / docstrings / example tutorials as needed.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.

CI

CI runs on self-hosted GPU runners and requires a maintainer to add the
run-ci label. Once labeled, every subsequent push re-triggers CI as
long as the label remains. Draft PRs are skipped even if labeled.

[zhaochenyang20]

  I'm good with the current docs. Do not need to change it.

[kevin85421]

Removed.