Repository navigation
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Hi! 👋 OpenAI announced a change that will break this project after February 26, 2027 (details below). This pull request updates the code so it keeps working.
It was prepared with the help of AI (YakBarber, a tool I'm building that updates code before API changes break it) and sent by me, once, to this repository only. If it isn't useful, feel free to close it — there won't be any follow-ups.
OpenAI is changing its API. whisper-1 is shut down; move transcription to gpt-transcribe
Deadline: February 26, 2027 (150 days from now). Details: https://developers.openai.com/api/docs/deprecations
What in this repo is affected
CHANGELOG.md:647— (documentation) This changelog entry historically describes issue [Bug] scores.json misdescribes the run: a hosted-API run records provider/region 'local', and a fallback-normalizer caveat is never written to warnings #184 and mentions whisper-1 only as an example value, it is prose that is never executed.psdn_sonar/benchmark/README.md:79— (documentation) This is an example JSON snippet in documentation showing a sample scores.json payload, not executable code.psdn_sonar/benchmark/submission.py:138— (documentation) This is a pydantic Field description string giving an example value for model_snapshot, it is metadata/help text, not code that talks to the provider.psdn_sonar/conf/backend/whisper.yaml:6— This config value is the literal OpenAI model name sent on POST /v1/audio/transcriptions for the whisper backend and must become gpt-transcribe.psdn_sonar/evaluators/single_speaker.py:93— (documentation) This is a docstring comment giving whisper-1 as an example of a provider-side model id, never executed.psdn_sonar/models/apis.py:68— This is the WhisperAPIModel constructor's default model parameter, which is sent as the 'model' field to OpenAI's transcription endpoint and must default to gpt-transcribe.psdn_sonar/models/apis.py:86— This conditional decides whether to send the legacy 'language' hint, which under gpt-transcribe must instead be sent via extra_body={'languages': [...]}.psdn_sonar/models/apis.py:87— This line sets params['language'] as a plain top-level kwarg for the OpenAI SDK call, which must become extra_body={'languages': [self.language]} per OpenAI's migration guide.psdn_sonar/models/base.py:89— (documentation) This is a docstring comment describing the provider_model_id attribute using whisper-1 as an example, not executable code.What this PR changes
Migrated the OpenAI Whisper adapter off the shut-down whisper-1 model. psdn_sonar/conf/backend/whisper.yaml now configures model: gpt-transcribe, and psdn_sonar/models/apis.py's WhisperAPIModel defaults to model="gpt-transcribe" and sends the language hint via extra_body={"languages": [self.language]} instead of the removed language= kwarg, per OpenAI's migration guide. No other call sites use whisper-1, response_format, timestamps, or the translations endpoint, so nothing else needed changes; the flagged documentation/docstring/README/changelog mentions of whisper-1 were left as-is since they're prose, not executable code. The test run still fails with the pre-existing "unrecognized arguments: --cov=..." pytest-cov error (present before my changes too), which is a test-configuration issue unrelated to this API migration and outside the files I'm allowed to edit.
Tests:⚠️ inconclusive — the tests failed the same way before this change in YakBarber's isolated test environment (often a setup difference, such as no network access), so they don't confirm this change (output below)
Test output
This pull request was generated by YakBarber using AI and checked against your tests. Please review before merging. Questions: reply to this pull request
Check the whole repository yourself:
npx yakbarber checklists every AI model and API it calls that is being retired, with the date and the company's own words. Free, read-only, runs locally (source).