Module rule_outcomes 

Local-only rule outcome telemetry.

Records when a rule is surfaced by recall (kind = 'recalled') so difflore memory can show “what rules are actually pulling weight” and rules show can include a use-count line.

Fix-acceptance telemetry already lives in fix_outcomes (richer schema with file path + applied/failed split); this table is just the recall-side ledger. Both are read together by the surfaces.

Data never leaves the device. Personal usage signal is not uploaded to cloud — cluster precision is derived cloud-side over consented aggregates only.

Structs

AcceptedFixEvidence

RecallCount

RecallSummary

RuleRecallInput

TopRecallEvidence

Constants

KIND_RECALLED

Functions

fix_accepted_count_for

How many fix_outcomes rows for this rule were accepted = 1 AND applied_ok = 1 within the window. Read here so the memory/show surfaces have a single import path for “rule outcome” reads.

latest_accepted_fix_for

latest_top3_recall_for

recall_count_for

Total recall count for a single rule over the last days days.

record_recalled

Insert one row per recalled rule. No-op when rule_ids is empty.

record_recalled_with_context

Insert recall rows with enough low-sensitive context to prove a value loop.

summary

Total local recall proof over the last days days.

top_recalled

Top-N rules by recall events within the last days days.