Support a more robust yet responsive SDLC within the Watson Assistant studio

Rather than suggest a solution (we have multiple ideas, I’m sure IBM and community would have many others) thought it better to state our problem.

Our needs:

• We need to speedily develop and promote conversation fixes to production to respond to problems with our dialogs

• We need to release new, larger conversations when they are QA’d. This takes longer than the quick fixes above.

• We need to be able to test large new conversations against test environments and data, both so we don’t trigger production events, and also for data privacy reasons – you know, the usual reasons.

Our reality:

• We have a separate skill for “Test” and “Prod” as we can only associate one webhook to each skill, so this is how we QA against test data and environments

• To make a quick fix we have to either make in prod skill, and then manually replicate the same changes in test skill, or make the change in test and manually replicate in the prod skill.

  • We cannot copy the whole test skill to prod as this would make new conversations that are still being developed “live” for customers

  • We cannot copy the prod skill back to the test skill as this would wipe out the conversations that are in development

  • So we have to do it twice and hope we remember/get it right. Which some times we won’t

Some solutions that could help include:

• Providing a compare tool to verify the differences between skills

• Allowing dynamic assignment of webhooks

• Supporting better tagging and replication of objects within a skill

Like I say, we don’t mind how its solved, but it is a real problem.