davinci-resolve-mcp 2.61.0 → 2.62.0

package/CHANGELOG.md

@@ -2,6 +2,110 @@
 
 Release history for the DaVinci Resolve MCP Server. The latest release is summarized in the root README; older entries live here to keep the README focused.
 
+## What's New in v2.62.0
+
+Community contribution by [@lukeashford](https://github.com/lukeashford)
+([#87](https://github.com/samuelgursky/davinci-resolve-mcp/pull/87), closing
+[#88](https://github.com/samuelgursky/davinci-resolve-mcp/issues/88)): server
+robustness for long operations plus a batch of timeline-tool ergonomics and
+correctness fixes. Every new option is opt-in; default behavior is unchanged.
+
+### Added
+
+- **Background jobs for long ops** — transcription, subtitle generation,
+  scene-cut/Dolby analysis, and timeline export/import can exceed the MCP
+  client's tool-window timeout. Passing `background=true` now returns a
+  `job_id` immediately; poll with `resolve_control(action="job_status")` or
+  list with `list_jobs` (both connection-free). Generic registry in
+  `src/utils/background_jobs.py`; workers run under the `resolve_busy` gate,
+  and finished jobs are pruned after an hour.
+- **`create_variant_from_ranges` pack mode** — `pack=true` butts clips
+  together at the end of each track (omits `recordFrame`), gap-free even when
+  source and timeline frame rates differ. Live-validated.
+- **`timeline set_current` by id/name** — accepts a stable `id`/`name`
+  selector (precedence id → name → index), not only the shifting 1-based
+  index.
+
+### Fixed
+
+- **Transport wedge** — synchronous tool bodies ran inline on the single
+  asyncio event-loop thread, so a blocking Resolve call (or the up-to-60s
+  launch wait) froze the whole server including the stdio read loop. Sync tool
+  bodies now run in a worker thread, serialized on a bridge lock so the
+  single-threaded scripting bridge is never entered concurrently; degrades to
+  inline behavior if the SDK shape changes.
+- **`create_variant_from_ranges` honesty** — results report actual placement
+  (`items[].placed` in both frame spaces, plus a `placement_mismatches`
+  count) instead of echoing the requested frames; `dry_run` resolves clip ids
+  and validates frame ranges so it fails on the same errors as the commit
+  path (and no longer leaves an orphan timeline on a bad id); the response
+  reports audio presence and warns on a video-only (silent) range list.
+- **`get_items` / `get_items_in_track`** — validate their track selector
+  (structured error instead of a bare `KeyError`), accept `index` or
+  `track_index`, and gained `action_help` entries.
+- **`apply_cuts`** — reports `skipped` cuts with reasons instead of silently
+  dropping non-applicable entries.
+- **`action_help`** — lists every valid action (not only the documented
+  subset) and distinguishes an unknown action (`UNKNOWN_ACTION`) from a
+  valid-but-undocumented one (`HELP_NOT_REGISTERED`).
+
+### Documentation
+
+- Corrected `create_variant_from_ranges` help: `clip_id` is a media-pool item
+  id (not a timeline-item id); `start_frame`/`end_frame` are SOURCE frames,
+  `end_frame` exclusive. Stated the timeline tool's frame-space convention
+  once, clarified `lift_range` ripple behavior on empty gaps,
+  `begin_run`/`end_run` batching, and the timeline-subtitle vs clip-level
+  transcript distinction.
+
+### Validation
+
+- Full offline suite: 1,483 tests green (70 new).
+- Transport offload, background jobs, and `create_variant_from_ranges`
+  placement/pack behavior were live-validated against Resolve Studio 21 by
+  the contributor; the threaded-dispatch SDK coupling was additionally
+  verified against the pinned mcp SDK (1.27) over a real stdio session.
+
+## What's New in v2.61.1
+
+The strata cleanup batch deferred from the v2.61.0 review — behavior-neutral
+consolidation and hot-path efficiency. No new actions or parameters.
+
+### Changed
+
+- **One clip resolver** — `strata.resolve_clip()` replaces the four drifting
+  per-module wrappers, and rides the pre-v9 auto-ingest fallback hoisted out
+  of deep_vision into `analysis_store.resolve_clip_uuid_ingesting()`: a clip
+  ref that resolves for `deepen` now resolves identically for every strata
+  action, including on older analysis roots.
+- **One float32 codec** — `pack_curve`/`unpack_curve` delegate to
+  `embeddings.pack_vector`/`unpack_vector` instead of duplicating the BLOB
+  convention.
+- **Decode once** — `strata_run` decodes the media file a single time when
+  several audio analyzers run (prosody + beat_grid previously each ran a full
+  ffmpeg decode); the shared `_audio_context` preamble also collapses their
+  duplicated require/resolve/decode blocks.
+- **Registry-derived capabilities** — `ANALYZERS` carries run function plus
+  requires/writes metadata; `capabilities()` derives from it, so adding an
+  analyzer is one entry.
+- **Query-layer caching** — word-find hits clustering in one clip unpack each
+  curve blob once, and `timeline_strata` resolves + bundles a source clip
+  reused across many placements once.
+- **Sargable event windows (schema v15)** — windowed `read_events` bounds its
+  span-overlap lower edge by the track's `MAX(duration_seconds)` (a b-tree
+  descent via the new `ix_events_clip_track_span` index) so
+  `ix_events_clip_track` range-seeks instead of scanning the track.
+- `backfill_words` iterates report blobs lazily with a transcription
+  prefilter instead of loading every blob into memory; `detect_breaths`
+  drops its unreachable non-numpy fallbacks.
+
+### Validation
+
+- Full offline suite: 1,431 tests green (13 new covering the shared resolver
+  + pre-v9 fallback, codec identity, decode-once, per-clip caching, long-span
+  window overlap, and the index range-seek query plan). No Resolve behavior
+  changed; live test not required.
+
 ## What's New in v2.61.0
 
 Perception strata — a timecoded track model over every analyzed clip, plus the

package/README.md

@@ -1,6 +1,6 @@
 # DaVinci Resolve MCP Server
 
-[![Version](https://img.shields.io/badge/version-2.61.0-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
+[![Version](https://img.shields.io/badge/version-2.62.0-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
 [![npm](https://img.shields.io/npm/v/davinci-resolve-mcp.svg?label=npm&color=CB3837)](https://www.npmjs.com/package/davinci-resolve-mcp)
 [![API Coverage](https://img.shields.io/badge/API%20Coverage-100%25-brightgreen.svg)](docs/reference/api-coverage.md)
 [![Tools](https://img.shields.io/badge/MCP%20Tools-34%20(341%20full)-blue.svg)](#server-modes)

package/install.py

@@ -35,7 +35,7 @@ from src.utils.update_check import (
 
 # ─── Version ──────────────────────────────────────────────────────────────────
 
-VERSION = "2.61.0"
+VERSION = "2.62.0"
 # Only hard floor: mcp[cli] requires Python 3.10+. There is no upper bound —
 # Resolve's scripting bridge loads into newer interpreters on recent builds
 # (Python 3.14 verified against Resolve Studio 20.3.2). Older Resolve builds

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "davinci-resolve-mcp",
-  "version": "2.61.0",
+  "version": "2.62.0",
   "description": "NPM bootstrapper for the DaVinci Resolve MCP Server.",
   "license": "MIT",
   "author": "Samuel Gursky <samgursky@gmail.com>",

package/src/granular/common.py

@@ -80,7 +80,7 @@ if not logging.getLogger().handlers:
         handlers=[logging.StreamHandler()],
     )
 
-VERSION = "2.61.0"
+VERSION = "2.62.0"
 logger = logging.getLogger("davinci-resolve-mcp")
 logger.info(f"Starting DaVinci Resolve MCP Server v{VERSION}")
 logger.info(f"Detected platform: {get_platform()}")