@withpica/mcp-server 2.94.0 → 2.95.0

package/CHANGELOG.md

@@ -9,6 +9,32 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 > matching entry here in the same commit. See the project's "npm publish
 > discipline" memory entry for the enforcement rationale.
 
+## [2.95.0] - 2026-07-24
+
+### Added
+
+- **`pica_audio_bulk_assign_suggested`** — attach ALL orphan audio whose
+  suggested match is high-confidence (ISRC-verified, non-ambiguous) in one
+  call, instead of looping `pica_audio_assign` per file (agent feedback
+  20f592ff: an agent hand-looped over a 74-file queue). Pairings are
+  recomputed server-side at execution time — stale suggestions can't apply,
+  dismissed pairs stay excluded, title-only matches are never bulk-attached.
+  Supports `dry_run: true` preview (same server-side classification, nothing
+  persisted — ADR-180 Rule 12).
+- **`pica_audio_dismiss_match`** — reject a suggested audio → recording
+  pairing so it never resurfaces in suggestions, the review queue, or the
+  bulk attach. The file stays attachable elsewhere; repeat dismissals are
+  no-op successes.
+
+### Changed
+
+- **`pica_recordings_query`** now surfaces the server's zero-result
+  `recovery_hint` on empty searches (shipped in sdk 3.8.0 — the hint was
+  built server-side but dropped on the wire; ops_issue 689a8e9a).
+- **`pica_audio_assign`** names a mis-slotted id on failure: when the value
+  passed as `work_id` is actually a recording id (or vice versa), the error
+  says so and names the right parameter — instead of an opaque FK failure.
+
 ## [2.94.0] - 2026-07-16
 
 ### Added