NLU

Installation prerequisites

This service need access to a volume on which the data will be saved and retrieved.
Currently, two drivers are supported : FILESYSTEM and AWS_S3, you can refer to the table below in order to set it up.

Also, make sure to enable the microservice in the provided helm charts as this microservice is disabled by default.

Environment variables

Name	Description	Default value
STORAGE_DRIVER	Storage driver used to store all bots related files. Available values : FILESYSTEM, AWS_S3	FILESYSTEM
GUNICORN_ACCESS_LOGFILE	If declared, the NLU will start logging HTTP logs to the specified folder, make sure the path is writable. Example : `/var/log/httpd-access.log`
TRAIN_ON_TALK_TIMEOUT	Upon calling the /talk API, timeout in seconds for which a 204 answer is returned if the bot training has not finished yet.	10
FILESYSTEM_LOCATION	Absolute path to the folder which will contain the bots	api/bots/datas/
AWS_LOCATION	Absolute path to the folder which will contain the bots	bots
AWS_ACCESS_KEY_ID	IAM access key ID
AWS_SECRET_ACCESS_KEY	IAM access key
AWS_STORAGE_BUCKET_NAME	S3 bucket name

Microservice testing

Once you configured and started the microservice (following the generic guide) you can verify everything is in order.

1. Create a bot:

   curl --location 'http://localhost:8080/nluv1/bots' \
   --header 'Content-Type: application/json' \
   --data '{
       "nlu": "snips"
   }'
   

   If successful, a bot object including multiple fields should be received, keep safe the value of the following fields : id, developer_token and client_token.

2. Talk with your bot:

   curl --location 'http://localhost:8080/nluv1/talk/{{id}}/en' \
   --header 'Authorization: Bearer {{client_token}}' \
   --header 'Content-Type: application/json' \
   --data '{
     "text": "Hello"
   }'
   

   As your bot has never been trained, it will train on empty data, it can take a few seconds to be ready. If your timeout variable is too low you might obtain a 204 response, it's ok you can retry in a few seconds.

   You should then receive an answer with the following format :

   {
       "input": "Hello",
       "intent": {
           "intentName": "Welcome Intent",
           "probability": 1
       },
       "slots": []
   }
   

3. Delete your bot:

   You are good to go, you can delete your assistant:

   curl --location --request DELETE 'http://localhost:8080/nluv1/bots/{{id}}/en' \
   --header 'Content-Type: application/json' \
   --header 'Authorization: Bearer {{developer_token}}' \
   --data ''
   

Congratulations your service is up and running!

Using from Prisme.ai

In order to use the microservice from the dedicated Prisme.ai app make sure to configure your prismeai-runtime instance with the following:

- name: APP_CONFIG_NLU_apiUrl
  value: 'http://apps-prismeai-nlu.apps.svc.cluster.local:8080' # Your URL might be different depending on namespaces, port and such...