Follow-up to server-26#115's window #4 measurement (CORRELATION_REVIEW_0913.md). The scenes map added by #132 (server-26#96/#114) captures transcript, incident_id, and corr_debug fields per scene, but not the scene's own extracted incident_type. That gap is why the newly-filed type-veto issue can't distinguish "extraction over-typed this specific scene" from "a sibling scene's type leaked into this scene's correlation ctx via the shared flat top-level field" -- both produce the same observable (final incident_type==None, veto=="type").
Add incident_type (and ideally severity, since _call_is_substanceless reads that too) to the per-scene map alongside the existing fields, so the next correlation measurement window can read a scene's own type directly instead of inferring it from ctx-at-decision-time.
Follow-up to server-26#115's window #4 measurement (`CORRELATION_REVIEW_0913.md`). The `scenes` map added by #132 (server-26#96/#114) captures `transcript`, `incident_id`, and `corr_debug` fields per scene, but not the scene's own extracted `incident_type`. That gap is why the newly-filed type-veto issue can't distinguish "extraction over-typed this specific scene" from "a sibling scene's type leaked into this scene's correlation ctx via the shared flat top-level field" -- both produce the same observable (final `incident_type`==None, veto=="type").
Add `incident_type` (and ideally `severity`, since `_call_is_substanceless` reads that too) to the per-scene map alongside the existing fields, so the next correlation measurement window can read a scene's own type directly instead of inferring it from ctx-at-decision-time.
Fixed by PR #152, merged to main. incident_type/severity now captured per-scene in the scenes map and surfaced in the debug dump — the next correlation measurement window can diagnose #138's type-veto question directly instead of guessing from a live dump.
Fixed by PR #152, merged to main. incident_type/severity now captured per-scene in the scenes map and surfaced in the debug dump — the next correlation measurement window can diagnose #138's type-veto question directly instead of guessing from a live dump.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Follow-up to server-26#115's window #4 measurement (
CORRELATION_REVIEW_0913.md). Thescenesmap added by #132 (server-26#96/#114) capturestranscript,incident_id, andcorr_debugfields per scene, but not the scene's own extractedincident_type. That gap is why the newly-filed type-veto issue can't distinguish "extraction over-typed this specific scene" from "a sibling scene's type leaked into this scene's correlation ctx via the shared flat top-level field" -- both produce the same observable (finalincident_type==None, veto=="type").Add
incident_type(and ideallyseverity, since_call_is_substancelessreads that too) to the per-scene map alongside the existing fields, so the next correlation measurement window can read a scene's own type directly instead of inferring it from ctx-at-decision-time.Fixed by PR #152, merged to main. incident_type/severity now captured per-scene in the scenes map and surfaced in the debug dump — the next correlation measurement window can diagnose #138's type-veto question directly instead of guessing from a live dump.