mathiasandClaude Sonnet 5.5 f25acebe4c feat(reconcile): merge two transcripts with a model choosing between disagreements ...

Base text stands except where the model picks the alternative's reading; trivial
disputes skip the model; the reading order alternates; the base's context goes to
the model. Fails soft: three consecutive model errors return the base text
unreconciled (none of the partial changes) with the reason; a verdict returned with
an error is ignored; a cancelled job is an error, not 'skipped'. Counts are exact,
the stored list of contested spots is capped at 200 and carries the second each
starts at. Mutation-checked on nine behaviours.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

2026-10-01 23:20:12 +02:00

..

adjudicate_test.go

test(reconcile): a number or word that merely starts with 1 or 2 is not an answer

2026-10-01 23:18:11 +02:00

adjudicate.go

feat(reconcile): LLM adjudicator with the measured prompt, plus an order-checked wrapper

2026-10-01 23:18:11 +02:00

diff_test.go

test(reconcile): word-level diff is minimal, tiles both inputs and is fast (red)

2026-10-01 23:15:44 +02:00

diff.go

feat(reconcile): minimal word-level diff (Myers) with a bounded edit distance

2026-10-01 23:16:31 +02:00

disputes_test.go

test(reconcile): finding disputes, resolving the trivial ones, and their context (red)

2026-10-01 23:16:55 +02:00

disputes.go

feat(reconcile): find disputes, resolve trivial ones without a model, give context

2026-10-01 23:17:21 +02:00

reconcile_test.go

test(reconcile): a verdict returned with an error is not used

2026-10-01 23:20:12 +02:00

reconcile.go

feat(reconcile): merge two transcripts with a model choosing between disagreements

2026-10-01 23:20:12 +02:00

tokens_test.go

test(reconcile): tokenising for word-level comparison (red)

2026-10-01 23:15:44 +02:00

tokens.go

feat(reconcile): tokenise for word-level comparison

2026-10-01 23:16:31 +02:00