SMS + NLU Demo

Install

These steps assume Python 3.4+, PortAudio and Speex are already installed. See the Mix.nlu Python sample app's README for more information.

You should also have a Twilio developer account, with a valid number that can send SMS.

Once you have all that, you can install your Python environment:

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Configure

1. Copy the credentials file and contact list templates:

   cp contacts.sample.json contacts.json
   cp creds.sample.json creds.json

2. Create an empty Mix.nlu model, then upload sms-model.trsx;

3. Publish a version of the model, and create an app config (context tag of your choice) pointing to it;

4. Copy the Mix.nlu credentials for the app config, and update creds.json;

5. Get your Twilio developer credentials, and update creds.json;

6. Edit contacts.json with the contacts you wish to use.

The LOG variable in nlu.py can optionally be set to False, in order to disable printing all WebSocket messages to the screen.

Run

Note: user_id should be set accordingly for each command

Data Upload

This command should be run first, with the actual list of contacts to be recognized.

python nlu.py --user_id=samueldg data_upload CONTACT contacts.json

Note: There is a short delay before the data upload actually kicks in for the recognition. Doing a couple of NLU requests usually does the trick.

ASR + NLU

python nlu.py --user_id=samueldg audio

Text + NLU

python nlu.py --user_id=samueldg text 'Say hi to myself'

Data Wipe

python nlu.py --user_id=samueldg data_wipe