v1.5.0

[1.5.0] - 2026-10-19

Added

• Import Muse native tool, read and skill evidence; report unmeasured usage and partial activity coverage instead of zeros

v1.4.2

[1.4.2] - 2026-10-19

Changed

• Credit intervention senders, preserve grouped verdicts and retain unknown actor reasons

v1.4.1

[1.4.1] - 2026-10-19

Changed

• Treat gh pr checks exit 8 as pending and correct existing ledgers

v1.4.0

[1.4.0] - 2026-10-19

Added

• Import Cue decisions and settled ContextMD reads read-only; report procedure reads by entry, harness, machine and AgentsMD version against session-wide holdout, with source and target gaps.

Changed

• sync now reads ContextMD Cue evidence by default alongside the T3 import; a failed Cue import records a gap without discarding other sync writes.
• health shows optional Cue gaps without failing; Cue status causes exit 4 only for schema_drift or unreadable_database.

Fixed

• Parse sectioned AgentsMD direction identity, Codex native durations and Read calls, and Grok MCP timing and outcomes. Retain read success state and count only successful procedure reads; replay older sources to repair stored evidence (#128).

v1.3.0

[1.3.0] - 2026-10-19

Added

• Discover Chromeria bare Codex and Grok homes by default; report trace truncation, unresolved Codex identity conflicts and unavailable health scans; recognize Claude request-replay metadata; faster availability and diagnose queries

Changed

• Judge moments JSON opening is null when unknown, and false for interrupts and rejections (was wrongly true)
• health exits 4 when its process scan is unavailable (previously it crashed)

Fixed

• health scans only processes its checks use, so T3 or a non-English locale cannot make it unavailable
• A file reached through two paths in one sync is imported under one source row

v1.2.6

[1.2.6] - 2026-10-15

Changed

• Count Claude queued inputs once without changing message identities

v1.2.5

[1.2.5] - 2026-10-15

Changed

• Keep schema-3 queries readable before occurrence metadata migration

v1.2.4

[1.2.4] - 2026-10-15

Changed

• Hydrate valid empty OpenCode assistant text without treating it as missing

v1.2.3

[1.2.3] - 2026-10-15

Changed

• Retain native Codex Extension web.search evidence and repair existing indexes

v1.2.2

[1.2.2] - 2026-10-14

Changed

• Index native conversation events and attribute actors by identity, keeping sidechain restrictions, Codex user text, and cleared OpenCode text aligned with those records. Codex orders every row by its complete-line position. Native ids keep the raw metadata ordinal, and a Codex privacy reimport of the same source updates that stored position from the first record without replacing the event or its verdict. A fork moves copied history by that file position on an append, a full replay, and a repair. A ledger imported before that chronology still moves on its stored metadata ordinals, and the replay then stores file positions. A repeat in the same source keeps that source's first position, and another file records another occurrence rather than moving the original. Competing response witnesses stay unplaced. A timeline orders the assembled rows from a spanning source before it applies a limit. Copied fork history keeps each native id, including two explicit ids that share a raw metadata ordinal, stores that record's file position, and retains the reads and fallback usage that attribute its responses. Only a ledger imported before privacy 9 rewrites those locators. A copied row matches the id its writer stored, including a file-position fallback, a historical ordinal fallback, and an id-less abort, and lookup still reads that historical id. The source is complete only when every retained locator is inside the file. A requested set stays unknown when the sources disagree about what preceded a shared identity, and each source sequence is kept. A timeline leaves sessions with no Codex order in timestamp order, applies a supported Codex order before the limit, and marks a cross-session order unknown. Opposite observations of the same native evidence count once, and that direction stays unknown. A failed replay rolls back so the next source cannot commit a partial replacement.