In @.github/workflows/publish_stable.yml:
- Around line 34-35: The workflow uses the deprecated ::set-output syntax in the
step that runs "echo \"::set-output name=version::$(python setup.py
--version)\"" with id "version"; either replace it with the environment file
method by writing the version to the GITHUB_OUTPUT file (e.g., echo
"version=$(python setup.py --version)" >> $GITHUB_OUTPUT) so later steps can
consume outputs, or remove the entire step (and its id "version") if that
computed version is not referenced by any subsequent step.
- Line 55: Replace the GitHub Action reference that currently uses the floating
ref "ad-m/github-push-action@master" with a specific, pinned release tag (e.g.,
the latest stable tag such as vX.Y.Z) so the workflow uses a fixed, versioned
commit; update the uses string in the workflow to that exact tag after verifying
the action's latest stable release on its repository.

In @.github/workflows/release_workflow.yml:
- Around line 3-20: The publish_alpha job currently runs on any pull_request
closed event; update the job-level condition for publish_alpha (job name:
publish_alpha) to only proceed when the PR was merged by adding an if that
checks github.event.pull_request.merged == true, and ensure downstream jobs that
depend on publish_alpha (e.g., publish_pypi, propose_release) either inherit
that merged check or also include an if: success() &&
github.event.pull_request.merged == true so the entire release pipeline cannot
run for unmerged PR closures.
- Around line 26-63: The "version" step uses the deprecated ::set-output syntax;
change it to write the output to the GITHUB_OUTPUT file (e.g., echo
"version=$(python setup.py --version)" >> $GITHUB_OUTPUT) and keep the step id
"version" so later steps can reference outputs as steps.version.outputs.version,
and replace the publish action reference uses:
pypa/gh-action-pypi-publish@master with a pinned release track (e.g.,
pypa/gh-action-pypi-publish@release/v1 or another stable tag) in the "Publish to
PyPI" step to avoid using the master branch.

In @.github/workflows/unit_tests.yml:
- Around line 4-10: Update the GitHub Actions workflow's pull_request branch
filter so PRs to the repo default branch trigger the job: in the workflow YAML
where the pull_request trigger is defined (the pull_request: branches: list),
add "dev" alongside "master" (or replace with the correct default branch name)
so the branches array includes both "master" and "dev" under the pull_request
key.
- Around line 11-14: The reusable workflow reference under the py_build_tests
job is pinned to a non-deterministic ref and the python_version is too low;
update the uses value (the uses field on the py_build_tests job) to a specific
tag or commit SHA instead of @master and change python_version from "3.8" to
"3.9" (or the minimum Python required by pinned dependencies) so the job is
deterministic and compatible with huggingface_hub==0.36.0.

In @ovos_audio_transformer_plugin_speechbrain_langdetect/__init__.py:
- Around line 3-8: The shim that sets torchaudio.list_audio_backends currently
returns an array with an empty string, which is not a valid backend name; change
the fallback assigned to torchaudio.list_audio_backends (the lambda defined in
__init__.py) so it returns an empty list [] to correctly indicate no available
backends instead of [''].

In @requirements.txt:
- Around line 1-8: The requirements file currently hard-pins
huggingface_hub==0.36.0 and leaves torchaudio and speechbrain unconstrained
which risks future breakage; verify whether the referenced SpeechBrain issue
exists and if the exact pin is required, otherwise relax huggingface_hub to a
range (for example huggingface_hub>=0.36.0,<1.0.0) and add conservative upper
bounds for torchaudio and speechbrain (e.g., torchaudio>=2.1.0,<3.0.0 and
speechbrain>=1.0.3,<2.0.0) in requirements.txt, then run the project’s test
suite and any integration flows that use SpeechBrain/huggingface_hub to confirm
compatibility before committing the change.