feat(gladia): add GladiaTranscriber integration for audio transcription - #3751
feat(gladia): add GladiaTranscriber integration for audio transcription#3751Aryan-Pardeshi wants to merge 4 commits into
Conversation
|
Hi @Aryan-Pardeshi, thanks a lot for your contribution! 🙏 We noticed that the Contributor License Agreement (CLA) check ( To get your PR reviewed, please sign the CLA via the link in the |
|
Thanks for signing the CLA, @Aryan-Pardeshi! 🎉 This PR is now ready for review again and the reviewer has been re-assigned. |
anakin87
left a comment
There was a problem hiding this comment.
Hello!
- I see that several files are missing (license, py.typed, ...). Please use the scaffold script to create a new integration as suggested here
- Have you had the chance to try this integration yourself with an API key? This is strongly recommended before submitting a new integration.
| documents = [] | ||
| with httpx.Client(timeout=self.timeout) as client: | ||
| for source in sources: | ||
| audio_url = self._upload_audio(client, source) |
Adds the files/registrations the create_new_integration.py scaffold script normally generates but were missing from this PR: LICENSE.txt, py.typed marker, tests/__init__.py, CI workflow (.github/workflows/gladia.yml), labeler.yml entry, coverage-comment workflow registration, and the root README table row. Also adds a real (currently-skipped) integration test that transcribes a committed sample audio file via the live Gladia API, gated on GLADIA_API_KEY like the other integrations in this repo. It will start running once a maintainer adds the secret.
|
Thanks for the pointer — I ran On point 2: I don't have a Gladia API key, so I can't personally verify this against the live API. I added a real integration test ( |
|
@Aryan-Pardeshi, I recommend checking whether Gladia provides free credits on the starter plan. Before committing to a deep review of the implementation, I expect the contributor to have properly tested their implementation locally. If you realize you cannot test it yourself, I think it's reasonable to close the PR. |
Closes #2246.
Adds a
gladiaintegration with aGladiaTranscribercomponent, modelled on Haystack'sRemoteWhisperTranscriberas suggested in the issue.Scoped to batch transcription over the Gladia v2 pre-recorded flow. The issue links Gladia's real-time product page while the design describes a batch component — flagging that explicitly so review does not stall on the ambiguity. Happy to follow up with streaming as a separate PR if that is what you want.
Unit tests mock
httpx; no API key or network needed.