121 lines
4 KiB
Python
121 lines
4 KiB
Python
import gevent.monkey
|
|
gevent.monkey.patch_all()
|
|
import base64
|
|
from email.mime.multipart import MIMEMultipart
|
|
from email.message import Message
|
|
from .model_map import get_model_for_lang
|
|
import json
|
|
import struct
|
|
import os
|
|
|
|
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
|
|
def handle_chunking():
|
|
request.environ['wsgi.input_terminated'] = 1
|
|
|
|
|
|
def parse_chunks(stream):
|
|
boundary = b'--' + request.headers['content-type'].split(';')[1].split('=')[1].encode('utf-8').strip() # super lazy/brittle parsing.
|
|
this_frame = b''
|
|
while True:
|
|
content = stream.read(4096)
|
|
this_frame += content
|
|
end = this_frame.find(boundary)
|
|
if end > -1:
|
|
frame = this_frame[:end]
|
|
this_frame = this_frame[end + len(boundary):]
|
|
if frame != b'':
|
|
try:
|
|
header, content = frame.split(b'\r\n\r\n', 1)
|
|
except ValueError:
|
|
continue
|
|
yield content[:-2]
|
|
if content == b'':
|
|
print("End of input.")
|
|
break
|
|
|
|
|
|
@app.route('/heartbeat')
|
|
def heartbeat():
|
|
return 'asr'
|
|
|
|
@app.route('/NmspServlet/', methods=["POST"])
|
|
def recognise():
|
|
stream = request.stream
|
|
|
|
access_token, lang = request.host.split('.', 1)[0].split('-', 1)
|
|
|
|
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')
|
|
|
|
body = {
|
|
'config': {
|
|
'encoding': 'SPEEX_WITH_HEADER_BYTE',
|
|
'language_code': lang,
|
|
'sample_rate_hertz': 16000,
|
|
'max_alternatives': 1,
|
|
'enableAutomaticPunctuation': True,
|
|
'enableSpokenPunctuation': True,
|
|
'model': get_model_for_lang(lang),
|
|
# '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 = []
|
|
if 'results' in result.json():
|
|
for result in result.json()['results']:
|
|
words.extend({
|
|
'word': x,
|
|
'confidence': str(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()
|
|
response_part = Message()
|
|
response_part.add_header('Content-Type', 'application/JSON; charset=utf-8')
|
|
|
|
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],
|
|
}))
|
|
else:
|
|
response_part.add_header('Content-Disposition', 'form-data; name="QueryRetry"')
|
|
# Other errors probably exist, but I don't know what they are.
|
|
# This is a Nuance error verbatim.
|
|
response_part.set_payload(json.dumps({
|
|
"Cause": 1,
|
|
"Name": "AUDIO_INFO",
|
|
"Prompt": "Sorry, speech not recognized. Please try again."
|
|
}))
|
|
parts.attach(response_part)
|
|
|
|
parts.set_boundary('--Nuance_NMSP_vutc5w1XobDdefsYG3wq')
|
|
|
|
response = Response('\r\n' + parts.as_string().split("\n", 3)[3].replace('\n', '\r\n'))
|
|
response.headers['Content-Type'] = f'multipart/form-data; boundary={parts.get_boundary()}'
|
|
return response
|
|
|