rebble-asr

asr.rebble.io: speech recognition for rebble

setup Google Speech-to-text credentials

In the Cloud Console, go to the Create service account page.

Create or select a project. (I'll use rebble-asr)

Enable the Cloud Speech-to-Text API for that project.

Create a service account. Set its permissions role only for Speech-to-Text

Download a private key as JSON. Save it as key.json

Setup Google Cloud Platform: Artifact Registry, Cloud Run

Enable Artifact Registry and Cloud Run

In Artifact Registry create a new repository. Set up the region and note it.

Set as Docker, choose a region, create repository

Build & Deploy

In Google Cloud Shell Terminal and Editor

git clone https://github.com/bdotq/rebble-asr

cd rebble-asr

in Dockerfile change $PORT to 443 if not already set

upload key.json to the asr folder

in asr/__init__.py change lang="en-US" to whatever language locale you wish to use. API_KEY is no longer used, SpeechClient will now refer to key.json. auth_req around line 58 has been commented out, its used to authenticate the user for the rebble.io service, but not needed for this

cd ..

docker build -t dictation .

With the REPOSITORY NAME you created in the REGION on Artifact Registry, modify the following command (where rebble-asr is your Google cloud project name):

docker tag dictation YOURREGIION-docker.pkg.dev/YOUR_REPO/rebble-asr/dictation

Then push to Artifact Registry, similarily modifying this command:

docker push YOURREGIION-docker.pkg.dev/YOUR_REPO/rebble-asr/dictation

Open up Artifact Registry

Navigate to the image digests that you just pushed [ YOURREGIION-docker.pkg.dev > YOUR_REPO > rebble-asr > dictation ]

Click on the most recent, Then click Deploy > Deploy on Cloud Run.

Set port to be 443 and the rest of defaults will be fine

After server is up running, you can use that is URL created (eg. https://dictation-########.a.run.app)

(Can also deploy with command gcloud run deploy dictation --image YOURREGIION-docker.pkg.dev/YOUR_REPO/rebble-asr/dictation:latest)

Use with Pebble

Log into auth.rebble.io/account

Under "Experimental Features for Developers", click 'I know what I'm doing'

In the text area paste the following with the URL modified with the server URL you got above

{
  "config": {
      "voice": {
          "languages": [
              {
                  "endpoint": "dictation-########.a.run.app",
                  "six_char_locale": "eng-USA",
                  "four_char_locale": "en_US"
              }
          ],
          "first_party_uuids": [
              ""
          ]
          }
      }
}