From 0d840bbb73986d9a8c026753248883d9d787d8ca Mon Sep 17 00:00:00 2001
From: Katharine Berry <katharine@kathar.in>
Date: Sat, 30 Jun 2018 15:20:01 -0700
Subject: [PATCH] Make it run in production.

---
 Procfile         |  1 +
 asr/__init__.py  | 80 +++++++++++++++++++++++++-----------------------
 requirements.txt | 19 ++++++++++++
 runtime.txt      |  1 +
 4 files changed, 63 insertions(+), 38 deletions(-)
 create mode 100644 Procfile
 create mode 100644 requirements.txt
 create mode 100644 runtime.txt

diff --git a/Procfile b/Procfile
new file mode 100644
index 0000000..2097dff
--- /dev/null
+++ b/Procfile
@@ -0,0 +1 @@
+web: gunicorn -k gevent -b 0.0.0.0:$PORT asr:app
diff --git a/asr/__init__.py b/asr/__init__.py
index e741ddd..b5a1050 100644
--- a/asr/__init__.py
+++ b/asr/__init__.py
@@ -1,14 +1,20 @@
+import gevent.monkey
+gevent.monkey.patch_all()
+import base64
 from email.mime.multipart import MIMEMultipart
 from email.message import Message
 import json
 import struct
+import os
 
-from flask import Flask, request, Response
-from google.cloud import speech
-
+import requests
+from flask import Flask, request, Response, abort
 
 app = Flask(__name__)
 
+AUTH_URL = "https://auth.rebble.io"
+API_KEY = os.environ['SPEECH_API_KEY']
+
 
 # We know gunicorn does this, but it doesn't *say* it does this, so we must signal it manually.
 @app.before_request
@@ -18,7 +24,6 @@ def handle_chunking():
 
 def parse_chunks(stream):
     boundary = b'--' + request.headers['content-type'].split(';')[1].split('=')[1].encode('utf-8').strip()  # super lazy/brittle parsing.
-    print("Boundary: " + boundary.decode('utf-8'))
     this_frame = b''
     while True:
         content = stream.read(4096)
@@ -28,7 +33,6 @@ def parse_chunks(stream):
             frame = this_frame[:end]
             if frame != b'':
                 header, content = frame.split(b'\r\n\r\n', 1)
-                print(content)
                 yield content[:-2]
             this_frame = this_frame[end + len(boundary):]
         if content == b'':
@@ -36,44 +40,45 @@ def parse_chunks(stream):
             break
 
 
-def parse_data():
-    boundary = b'--' + request.headers['content-type'].split(';')[1].split('=')[1].encode('utf-8').strip()  # super lazy/brittle parsing.
-    parts = request.data.split(boundary)
-    for part in parts:
-        if part == b'':
-            continue
-        yield part.split(b'\r\n\r\n', 1)[1][:-2]
-
-
 @app.route('/NmspServlet/', methods=["POST"])
 def recognise():
-
-    client = speech.SpeechClient()
     stream = request.stream
+
+    access_token, part1, part2 = request.host.split('.', 1)[0].split('-', 3)
+    lang = f"{part1}-{part2.upper()}"
+
+    auth_req = requests.get(f"{AUTH_URL}/api/v1/me/token", headers={'Authorization': f"Bearer {access_token}"})
+    if not auth_req.ok:
+        abort(401)
+
     chunks = iter(list(parse_chunks(stream)))
     content = next(chunks).decode('utf-8')
-    print(content)
 
-    config = speech.types.RecognitionConfig(
-        encoding='SPEEX_WITH_HEADER_BYTE',
-        language_code='en-US',
-        sample_rate_hertz=16000,
-    )
-    print('beginning request')
-    responses = client.streaming_recognize(
-        config=speech.types.StreamingRecognitionConfig(config=config),
-        requests=(
-            speech.types.StreamingRecognizeRequest(audio_content=struct.pack('B', len(x)) + x)
-            for x in chunks))
-    print('finished request')
+    body = {
+        'config': {
+            'encoding': 'SPEEX_WITH_HEADER_BYTE',
+            'language_code': lang,
+            'sample_rate_hertz': 16000,
+            'max_alternatives': 1,
+            # 'metadata': {
+            #     'interaction_type': 'DICTATION',
+            #     'microphone_distance': 'NEARFIELD',
+            # },
+        },
+        'audio': {
+            'content': base64.b64encode(b''.join((struct.pack('B', len(x)) + x for x in chunks))).decode('utf-8'),
+        },
+    }
+    result = requests.post(f'https://speech.googleapis.com/v1/speech:recognize?key={API_KEY}', json=body)
+    result.raise_for_status()
+
     words = []
-    for response in responses:
-        if response.results:
-            for result in response.results:
-                words.extend({
-                                 'word': x,
-                                 'confidence': result.alternatives[0].confidence
-                             } for x in result.alternatives[0].transcript.split(' '))
+    if 'results' in result.json():
+        for result in result.json()['results']:
+            words.extend({
+                             'word': x,
+                             'confidence': result['alternatives'][0]['confidence']
+                         } for x in result['alternatives'][0]['transcript'].split(' '))
 
     # Now for some reason we also need to give back a mime/multipart message...
     parts = MIMEMultipart()
@@ -83,6 +88,7 @@ def recognise():
     if len(words) > 0:
         response_part.add_header('Content-Disposition', 'form-data; name="QueryResult"')
         words[0]['word'] += '\\*no-space-before'
+        words[0]['word'] = words[0]['word'][0].upper() + words[0]['word'][1:]
         response_part.set_payload(json.dumps({
             'words': [words],
         }))
@@ -96,10 +102,8 @@ def recognise():
             "Prompt": "Sorry, speech not recognized. Please try again."
         }))
     parts.attach(response_part)
-    print(parts.as_string())
 
     response = Response(parts.as_string().split("\n", 3)[3])
     response.headers['Content-Type'] = f'multipart/form-data; boundary={parts.get_boundary()}'
-    response.headers['Connection'] = 'close'
     return response
 
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..2df9200
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,19 @@
+certifi==2018.4.16
+chardet==3.0.4
+click==6.7
+Flask==1.0.2
+gevent==1.3.4
+greenlet==0.4.13
+gunicorn==19.8.1
+idna==2.7
+itsdangerous==0.24
+Jinja2==2.10
+MarkupSafe==1.0
+pyasn1==0.4.3
+pyasn1-modules==0.2.2
+pytz==2018.4
+requests==2.19.1
+rsa==3.4.2
+six==1.11.0
+urllib3==1.23
+Werkzeug==0.14.1
diff --git a/runtime.txt b/runtime.txt
new file mode 100644
index 0000000..1935e97
--- /dev/null
+++ b/runtime.txt
@@ -0,0 +1 @@
+python-3.6.6