Verify a rewritten article actually fixes its retrieval misses after publish

It closes the loop on article fixes. When an article is published or edited in Intercom, the workflow replays the real user phrasings that previously failed to retrieve it and checks whether the updated content now ranks first. This proves the rewrite worked instead of assuming it did.

Use it right after you ship article rewrites driven by your retrieval audits. It prevents 'fixed' articles from silently still missing the queries they were supposed to catch, and it keeps your Linear board honest.

1. 1An Intercom webhook fires when an article is published or updated.
2. 2Load that article's historical missed phrasings from Postgres.
3. 3An OpenAI step re-runs retrieval scoring for each phrasing against the updated content.
4. 4A logic branch checks whether the article now ranks top for every phrasing.
5. 5If all pass, mark the Postgres record resolved and the loop ends.
6. 6If any still miss, reopen the Linear ticket with the remaining failing phrasings and current ranks.