From bee2b2fdbb669b71893de33b6f5eaecfdb6058ff Mon Sep 17 00:00:00 2001 From: Nubenetes Bot Date: Sun, 14 Jun 2026 09:57:53 +0200 Subject: [PATCH] docs(automation): expand recovery documentation with finalization steps --- README.md | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index 0fee3ce9..c8c3e81d 100644 --- a/README.md +++ b/README.md @@ -965,7 +965,11 @@ When processing massive datasets (e.g., full AI re-evaluations of 18k+ links), w * **Run Workflow**: Select the `develop` branch. * **restore_cache**: Set to `true` (Downloads the last saved state). * **force_reevaluate**: Set to `false` (Ensures the AI only processes links that are still pending/missing metadata). -3. **Efficiency**: This strategy ensures zero token waste and zero loss of processing time, allowing the system to reach 100% coverage across multiple successive runs if necessary. +3. **Finalization and Promotion**: Once the AI recovery run completes successfully (which will be much faster as it only processes pending items): + * **Manual Publisher Trigger**: If the automated chain does not trigger, manually run the `04.1. V2 Publisher` workflow. + * **Automated PR**: The Publisher will automatically create or update the Pull Request from `develop` to `master`. + * **Human Review**: The final step remains a manual review and merge of the PR by the repository owner to deploy to production. +4. **Efficiency**: This strategy ensures zero token waste and zero loss of processing time, allowing the system to reach 100% coverage across multiple successive runs if necessary. ---