21 lines (14 loc) · 854 Bytes

Changelog

v0.2.0

Stronger core behavior

added explainable site_candidates output instead of only returning one chosen domain
added review verdicts (ready, review_required, blocked) to make human review explicit
added ranked top_contact_candidates with reasons and source attribution
kept confidence scoring, but turned it into a clearer decision surface

Fewer embarrassing failures

generate_outreach.py now refuses to draft by default when the dossier is not ready
added --allow-review-required override for conscious manual use
preserved support for weak-but-official examples while making the risk visible

Proof

expanded tests to cover review verdicts, contact explanations, and CLI refusal/override behavior
refreshed examples to show both a weak-contact review case and a ready-to-draft case