multilingual-chatbot

Based on a Think 2018 lab session:

Watson Assistant + Watson Language Translator to create a multilingual chatbot

Setup

Verify that your system satisfies the dependencies below:

Make sure the Cloud Functions plug-in is installed

bx plugin install cloud-functions

Create a python environment with conda

conda env create -f environment.yml

Activate the environment

source activate watson

1. Create Watson services

Watson Language Translator

From the dashboard, click on Create resource and search for Language Translator.

Provision an instance of Language Translator under the Lite plan.

Create a set of credentials by clicking Service credentials > New credential > Add.

Grab the value for apikey of the newly created IAM credential. Store this in the file etc/params.json under "translator_apikey".

Watson Assistant

From the dashboard, click on Create resource and search for Watson Assistant.

Provision an instance of Watson Assistant under the Lite plan.

Create a set of credentials by clicking Service credentials > New credential > Add.

Grab the value for apikey of the newly created IAM credential. Store this in the file etc/params.json under "assistant_apikey".

Click Launch Tool.

Create a new skill by clicking on Create a Skill > Create new > Import skill > and upload car_workspace.json with the option Everything.

From the Skills page of the tool, click the three dots, then View API Details to get the Workspace ID. Store this in the file etc/params.json under "assistant_workspace_id".

Once Watson has finished training, you may test out interacting with the assistant by clicking on Try it on the right side of the page.

Challenge: try editing your workspace to customize the flows and responses.

2. Deploying your Cloud Function

Configure the Bluemix CLI

bx login

Target the organization and space that contains your services (follow the prompts)

bx target --cf

Deploy the action

Make sure you are in the multilingual-chatbot directory.

Run the deployment script to package and update your cloud function

./deploy_cloudfunction.sh

Binding your Watson services

Bind the parameters you have defined to the deployed cloud function

bx wsk action update translator --param-file etc/params.json

Verify that these values are available as parameters to your cloud function

bx wsk action get translator parameters

Send a few sample requests

bx wsk action invoke translator --result
bx wsk action invoke translator --result --param text "hi there"
bx wsk action invoke translator --result --param text "hola amigo"

3. Run a full conversation

Get the namespace for your deployed cloud function

bx wsk action get translator namespace

Run the program. Experiment with switching between languages mid-conversation

python main.py --namespace <namespace>

4. Integrate with an existing application

Make a standard HTTP POST request to interact with the cloud function from any application

curl -X POST -H "Content-Type: application/json" -d '{"text": "bonjour"}' https://openwhisk.ng.bluemix.net/api/v1/web/<namespace>/default/translator.json

You will see a response like this

{
  "context": "<content>",
  "intents": "<intents>",
  "language": "<language>",
  "message": "<message>",
  "output": "<output>"
}

On subsequent requests, make sure to pass that context variable back in, to continue the conversation where you left off

curl -X POST -H "Content-Type: application/json" -d '{"text": "jouer de la musique", "context": "<context>"}' https://openwhisk.ng.bluemix.net/api/v1/web/<namespace>/default/translator.json

Check out this demo for a full example

https://multilingual-chatbot.mybluemix.net/

with-watson / multilingual-chatbot

readme