Dedupe call feature requests into an Airtable tracker

For every call transcript it isolates the explicit feature requests a customer made, then checks an Airtable tracker for an existing matching request. If one exists, it increments the request count and appends the new account and quote; if not, it creates a fresh row. The result is a deduplicated, vote-counted feature ledger.

Use it when feature requests pile up as duplicate one-liners across hundreds of calls and you need a single tracker that shows true demand per request rather than raw mention count.

1. 1A finished call transcript starts the run.
2. 2The model extracts discrete, explicit feature requests from the conversation.
3. 3For each request it semantically searches the Airtable tracker for an equivalent existing entry.
4. 4A branch decides match versus new.
5. 5On a match, it updates the row: bump the count, add the requesting account and a supporting quote.
6. 6On no match, it creates a new tracker row with the request, source account, and first quote.