peptiderepo 2300fbf1c5 fix: scope dedup to batch + recent 7 days, not all-time history ...

Dedup now has two narrow scopes instead of checking all posts ever:

1. Intra-batch: prevents proposing two articles on the same topic
   within a single pipeline run (keyword overlap ≥60%)
2. Recent posts: checks only posts from the last 7 days to avoid
   back-to-back similar content

Older articles are intentionally NOT checked — revisiting a topic
weeks later is fine. This prevents the dedup from becoming
increasingly aggressive as the blog grows, which was the root cause
of "no viable article ideas found" with 125 collected items.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-04-19 02:53:59 +07:00

9.4 KiB

Executable File

Raw Blame History

View Raw