verbio-technologies / python-verbio-speech-center Goto Github PK

Description: [Suggestion] Secure Host Argument

Motivation and Context

Currently, the host argument is described as follows:

--host HOST, -H HOST The URL of the host trying to reach (default: speechcenter.verbio.com:2424)

Though the parameter exhibits a default value that serves the purpose of an example as well, any user could be blindly encourage to add an schema prefix. If said schema is provided to the gRPC library (e.g. https://speechcenter.verbio.com:2424), a misleading error is encountered stating that the DNS is unable to find/reach the host.

In order to avoid potential misuses, I would suggest using an URL parsing library (e.g. urllib.parse) to precisely gather the expected contents from the host parameter.

As a code example:

>>>import urllib.parse

# From either of the following examples, take SplitResult.hostname and SplitResult.port to compose the real gRPC host
>>>urllib.parse.urlsplit('speechcenter.verbio.com:2424')
SplitResult(scheme='', netloc='', path='speechcenter.verbio.com:2424', query='', fragment='')
>>> urllib.parse.urlsplit('http://speechcenter.verbio.com:2424')
SplitResult(scheme='http', netloc='speechcenter.verbio.com:2424', path='', query='', fragment='')
>>> urllib.parse.urlsplit('https://speechcenter.verbio.com:2424')
SplitResult(scheme='https', netloc='speechcenter.verbio.com:2424', path='', query='', fragment='')

Hello @jordi-adell, @anikocharyan, Verbio development team:
I am trying to test how Verbio ASR behaves for numbers and alphanumeric speech with this client, and would like to feedback an issue and ask some questions:

1. When running recognizer_stream.py with '--formatting', it still gives non-formatted result. It seems the following code causes the issue:
   parser.add_argument('--formatting', '-f', help='', required=False, default=False, action='store_false')

2. After fixing this locally, I can get text like "dee ef Dee 6 9 8 6 9". Seems 'dee' means D, 'ef' means F --- is this understanding correct? If correct, is there a full mapping of the words (like 'dee') to letters (like 'd')?

Description: [Refactor] Misleading Token Argument

Motivation and Context

Currently, the token argument is described as follows:

--token TOKEN, -t TOKEN A string with the authentication token

Despite the description, the current behaviour of the CLI client involves reading the authentication token from a text file, expecting a filename as the content of the --token argument. Therefore, I believe there is a mismatch between the parameter description and the expected contents which should be addressed.

Consequently, I propose either changing the description to a better fitting proposition or altering the behaviour of the CLI client with respect to the argument, expecting the token itself as the content.