Django + GCP (GAE, Speechi-to-Text API, Cloud Storage, Cloud Firestore) で文字起こしアプリをデプロイする。 Django gcp

目的

• Django アプリを GAE でデプロイする。
• Cloud Speech-to-Text のクライアント・ライブラリー を使用する。
• Cloud Storage と Cloud Firestore にデータを保存する。

アプリケーションの全体像

コード

動作確認

コードの抜粋

def speech(request):
    import os
    from google.cloud import storage
    from google.cloud import speech
    from google.cloud import firestore
    import re

    # for local
    key_path = './app1/key/xxx.json'
    os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = key_path
    CLOUD_STORAGE_BUCKET = 'xxx'
    CLOUD_STORAGE_BUCKET2 = 'xxx'

    # for deploy
    #CLOUD_STORAGE_BUCKET = os.environ['CLOUD_STORAGE_BUCKET']
    #CLOUD_STORAGE_BUCKET2 = os.environ['CLOUD_STORAGE_BUCKET2']

    path = 'gs://' + CLOUD_STORAGE_BUCKET + '/'

    if request.method == 'POST':
        form = SpeechForm(request.POST, request.FILES)
        if form.is_valid():
            blob_name = form.cleaned_data['name']
            sp = request.FILES['speech']
            #speech = form.cleaned_data['speech']

            # Cloud Storage            
            client_storage = storage.Client()
            bucket = client_storage.get_bucket(CLOUD_STORAGE_BUCKET)            
            blob = bucket.blob(blob_name)            
            blob.upload_from_string(
                    sp.read(),
                    )

            # Cloud Speech
            client_speech = speech.SpeechClient()
            gcs_uri = 'gs://'+CLOUD_STORAGE_BUCKET+'/'+blob_name

            audio = speech.types.RecognitionAudio(uri=gcs_uri)
            config = speech.types.RecognitionConfig(
                #encoding=speech.enums.RecognitionConfig.AudioEncoding.FLAC,
                encoding=speech.enums.RecognitionConfig.AudioEncoding.LINEAR16,
                sample_rate_hertz=16000,
                language_code='ja-JP')

            operation = client_speech.long_running_recognize(config, audio)
            response = operation.result()

            text = response.results[0].alternatives[0].transcript

            # Cloud Storage            
            client_storage = storage.Client()
            bucket = client_storage.get_bucket(CLOUD_STORAGE_BUCKET2)            
            blob = bucket.blob(blob_name)            
            blob.upload_from_string(
                    text,
                    )

            sales = 0
            s = re.search('売上高', text)
            if s is not None:
                text2 = text[s.end():s.end()+20]
                s2 = re.search('[0-9]+\S円', text2)
                sales = s2.group()

            o_income = 0
            s = re.search('営業利益', text)
            if s is not None:
                text2 = text[s.end():s.end()+20]
                s2 = re.search('[0-9]+\S円', text2)
                o_income = s2.group()

            n_income = 0
            s = re.search('純利益', text)
            if s is not None:
                text2 = text[s.end():s.end()+20]
                s2 = re.search('[0-9]+\S円', text2)
                n_income = s2.group()

            # Cloud Firestore
            firestore_client = firestore.Client()
            doc_ref = firestore_client.collection('texts').document(blob_name)

            sales = sales
            o_income = o_income
            n_income = n_income

            doc_ref.set({
                'sales': sales,
                'o_income': o_income,
                'n_income': n_income,                           
            })

            d = {
                    'text': text,
                    'sales': sales,
                    'o_income': o_income,
                    'n_income': n_income,
                }

            return render(request, 'app1/text.html', d)

    else:
        form = SpeechForm()   

    d = {
        'form': form,
    }

    return render(request, 'app1/speech.html', d)