Nightly duplicate-cluster digest for Front inboxes

Once a night this workflow scans all open Front conversations from the trailing 24 hours, groups them into clusters of likely duplicates, and sends support leads a ranked digest in Slack showing each cluster, its size, and a representative subject line.

Use it when you'd rather review duplicates in one batch than get interrupted per-ticket. Good for teams that triage in a morning standup and want a single "here's what to merge today" list.

1. 1A scheduled trigger runs every night after business hours.
2. 2The workflow fetches open conversations updated in the last 24 hours from Front.
3. 3OpenAI embeds each conversation's subject and opening message.
4. 4A Postgres step stores the embeddings and runs a similarity grouping query to form clusters above threshold.
5. 5A logic step drops singleton clusters and ranks the rest by member count.
6. 6OpenAI writes a one-line label for each cluster describing the shared issue.
7. 7The workflow posts the ranked digest to a Slack channel with deep links into each Front conversation.