[ShivamSharma43]

Problem

google.STT.update_options assigns languages, detect_language, interim_results, punctuate, spoken_punctuation and profanity_filter first, and only then checks that the adaptation fits the model's API version:

from google.cloud.speech_v1.types import resource as resource_v1

stt = google.STT(model="chirp_3", languages="en-US")
stt.update_options(languages="fr-FR", punctuate=False, adaptation=resource_v1.SpeechAdaptation())
# ValueError: adaptation must be cloud_speech_v2.SpeechAdaptation for v2 models, got SpeechAdaptation

The call raises, but the STT now has languages == ["fr-FR"] and punctuate == False. Streams opened afterwards use them, although the caller was told the update failed. A model switch that the current adaptation doesn't fit leaves the same partial state (for example, update_options(model="long", languages="de-DE") on a STT that holds a v2 adaptation).

Fix

update_options now resolves the target API version and validates the adaptation before it assigns anything. Nothing else changes: the same calls are accepted, the error messages are the same, and open streams are updated only after the checks pass, as before.

Tests

Two tests added to tests/test_plugin_google_stt.py:

• test_update_options_rejected_adaptation_changes_nothing: a v1 adaptation on chirp_3, passed with every option assigned before the check, leaves all of them unchanged.
• test_update_options_rejected_model_switch_keeps_languages: switching to a v1 model while a v2 adaptation is set leaves the languages, punctuate and the model unchanged.

Both fail on main.

pytest tests/test_plugin_google_stt.py --plugin google    # 27 passed
ruff check / ruff format --check on the changed files
mypy -p livekit.plugins.google                              # Success: no issues found

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.