I just finished an demo project posted by Christophe Laleveé on
developerWorks. It's called 'Build an enhanced IT help desk chatbot on
IBM i with Watson Conversation' and the link (today) is
In a nutshell, it's a chatbot that runs on Node-Red. It listens to a
Slack channel, sends questions posted there to Watson Conversation,
which extracts out the 'intent' of the question. Watson sends the
'intent' back to Node-RED, which then takes appropriate action.
The demo handles:
1) Lost/forgot password. Sends a text message with an authentication
code, resets the *USRPRF to a one-time password.
2) Can't connect. Similar flow to above.
3) CPU load. Reports back the CPU%.
4) Disk usage. Reports back the %used.
One of the actions is to interact with DB2 by forming SQL statements to
do run DB2 services like "SELECT NO_PASSWORD_INDICATOR FROM
QSYS2.USER_INFO WHERE AUTHORIZATION_NAME = UCASE('" +
msg.payload.context.userprf + "')"
It was pretty easy to set up, and strangely easy to modify. My first
change was to add a new intent in Watson to let Watson work out if the
user needs to know how / what to type in order to get help. An empty
chat box can be intimidating. I'd say it took me about 2 minutes to add
a new intent and dialogue to implement that, and it was available in
real time. No compile / deploy, just add the intent and it's there.
Most of the demo can be exported as JSON text / code, but there is some
set-up that needs a GUI. My decades as an RPG programmer are yelling at
me for not having /all/ the source code in my possession, but I'm trying
to get over that feeling.
IBM i 7.3, TR3.