Which account?

Demo. Uses made-up sample data from assets/, connects to no real system, and changes nothing. It's a reference for how to use Elicitly: adapt it to your own data and systems.

"Log this against Harbor" matches three accounts. Rather than guess, the agent lists the candidates with enough detail to tell them apart and lets a human choose. The choice is a real radio button on the review page, and only a valid account id can come back.

Steps

1. Read assets/matches.json. It is a complete elicit_selection input.
2. Call elicit_selection with that object as-is.
3. Give the user the reviewUrl as a link, and say the choice is waiting there.
4. Call elicit_await with the elicitationId. While it returns pending, call it again. Stop when the status is anything else.
5. If the status is submitted or declined, call elicit_result:
   • decision: "submitted": look up the chosen id in selectedIds among the options and name the account. This demo logs nothing; say "In a real skill, the call notes would be logged to this account here."
   • decision: "declined": none of the accounts was right. Ask the user for another detail (city, contact name) to narrow the search. For expired or canceled, say no choice was made and offer to ask again.