- Reorder: send Discord FIRST; commit CHANGELOG + version bump ONLY on a successful announce, so a failed send never consumes a version. - Raise max_tokens for the deepseek-v4-flash reasoning model (6000/4000) so reasoning+content fit and content is never empty. - Guard against empty content (don't POST an empty Discord message). - Add workflow_dispatch (manual re-fire) with an optional base ref input. |
||
|---|---|---|
| .. | ||
| workflows | ||
| FUNDING.yml | ||