Tracking Speech API development.

[daspecster]

These are the items required to reach feature completion.

• Use Transcript object for sync and async requests. (Use Transcript object in sync and async. #2613)
• GAPIC Sync Recognize (Add Speech GAPIC for sync_recognize. #2615)
• GAPIC Async Recognize(Clean up async docs and unify for future GAPIC async. #2638, Add speech async gapic #2663)
• Streaming recognize (Add Speech Streaming API. #2523, Add _make_streaming_request, formerly _make_streaming_config. #2640,Add _stream_requests() for managing speech streaming configuration #2644, Add speech streaming recognition. #2680)
• Add stability information to streaming interim_results (Add stability to speech API results. #2702)
• Add sync_recognize system tests (Add Speech sync recognize system test. #2547)
• Add async_recognize system tests (Add speech async JSON/REST system tests. #2648)

Cleanup

• Update SpeechContext in tests. See: Update speech async_recognize tests #2700. (Fix bug with speech streaming speech_context. #2717)
• Add PyPI badges (Add PyPI badges for speech #2716)
• Rename Transcript to Alternative. (Rename Transcript to Alternative. #2666)
• Test cleanup (Speech unit tests cleanup. #2676)
• Run skipped tests(Add speech async JSON/REST system tests. #2648 (comment))
• Move sample to Client property. See: Move speech methods to Sample. #2516
• Remove warning about google-cloud-speech not yet released. See: Add message about API not being available yet. #2620 (Remove speech package not yet released warning. #2732)

Future Updates:

• Clearer way to send stream to Sample.content (Decide how to handle stream vs bytes content in Speech streaming_recognize(). #2708)
• Move result access to Operation handler Clean up async docs and unify for future GAPIC async. #2638 (comment)
• Add snippet tests (optional for release)

[daspecster]

@dhermes I commented on your gist from today but I'll copy it here as well...
Ref: https://gist.github.com/dhermes/09c964d6d27003ae817b650424fda7c3

---

Thank you for this!

I have another question.
As you saw in your output here, the final result has a response attribute.

In the sample the way that you unpack the response is via

response = cloud_speech_pb2.AsyncRecognizeResponse()
operation.response.Unpack(response)

But that has to be done after the Operation is completed.
How would we handle that in google-cloud-python?

The point of async is not to block right? If our lib has to wait until the operation is complete before the data can be unpacked then it would block right?

One thought I had was to add a helper of some kind that unpacks the data for us.

import time
from google.cloud import speech
from google.cloud.speech.encoding import Encoding

client = speech.Client()
sample = client.sample(source_uri='gs://ferrous-arena-my-test-bucket/sample.raw', encoding=Encoding.LINEAR16, sample_rate=16000)

operation = client.async_recognize(sample, max_alternatives=2)

retry_count = 10
while retry_count > 0 and not operation.complete:
    retry_count -= 1
    time.sleep(1)

    operation.poll()  # API call

for result in speech.unpack_async(operation):  # The helper `unpack_async`.
        print('Result:')
        for alternative in result.alternatives:
            print(u'  ({}): {}'.format(
                alternative.confidence, alternative.transcript))

But otherwise I'm not sure how to get the data without blocking unless we do some kind of weird pubsub design.

[daspecster]

Scratch that...we could just parse it in Operation.poll() right? I think I would have to either override google.cloud.core.operation.Operation.poll or I could add some kind of polling proxy method.

[dhermes]

@daspecster It's not up to use to get the data, just give the user the Operation and let them decide how to poll.

[fehrenbacher]

Is this available yet? I've installed google-cloud 0.20.0, but can't find the google.cloud.speech.client.Client class. I've also found this page for a separate google-cloud-speech library, but pip can't find any actual releases there. The documentation sure makes it sound like this is available to use now...

[gw00207]

@fehrenbacher yes, that documentation is confusing. hopefully a message is added soon to explain: #2620

[daspecster]

@fehrenbacher it's not released yet, however if you install from source you can play with the API.

As you can see in the first link, it's at the "Planning" stage. That package is a place holder.

The documentation message hasn't gone up because there hasn't been a release yet. We're working on a better release strategy as well.

[fehrenbacher]

K thanks for clearing that up!

[daspecster]

@fehrenbacher it's my mistake with having the docs out there. I think I have a better process for next time. If you decide to install from source and play with it, please let me know if you run into any issues!

I have a handful of things I'm working to resolve this week and then I'm hoping we will be able to release it.

See: https://github.com/GoogleCloudPlatform/google-cloud-python/issues?q=is%3Aissue+is%3Aopen+label%3Aspeech

[gw00207]

@daspecster nice :-) looking forward to it!

[daspecster]

Closing this since everything is complete and I opened #2842 to track the last few little changes.