davinci-resolve-mcp 2.121.0 → 2.123.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -50,10 +50,13 @@ window. `render.verify_output` covers the container-level checks.
50
50
  | Custom start timecode | `spec.startFrame` / `preserveStartTimecode` | v2.117 |
51
51
  | Timeline markers | `spec.markers` (16 colors, notes, durations, customData) | v2.118 |
52
52
  | Turnover markers | EDL `* LOC:` locators + OTIO markers → authored | v2.119 |
53
+ | TC-bearing sources (AAF route) | native clip capture (`MediaStartTime`), channel-leg merge | v2.120–2.122 |
53
54
  | Fusion titles | `elements: [{type:'title', text}]` — **21-gen hosts only** | v2.108 |
54
55
 
55
56
  `assemble_from_interchange` drives the same engine from an EDL / OTIO /
56
- FCP7-XML / AAF plus a `sourceMap`, and returns an honesty ledger
57
+ FCP7-XML / AAF plus a `sourceMap` all four formats are route-proven
58
+ end-to-end (parse → assemble → import → measured frames and RMS) — and
59
+ returns an honesty ledger
57
60
  (`authoredTransitions`, `droppedTransitions` with reasons, `authoredRetimes`,
58
61
  `flattenedRetimes`, `authoredAudioEvents`, `upperTrackCutsVideoOnly`).
59
62
 
@@ -98,7 +101,10 @@ Everything else stays in `droppedTransitions` with the reason.
98
101
  ## Verification checklist for a delivered .drt
99
102
 
100
103
  1. `timeline.import_timeline_checked` — expect `linked == total` for media
101
- (generators legitimately count as offline).
104
+ (generators legitimately count as offline). For `.drt`/`.drp` it also
105
+ cross-checks the files items ACTUALLY link against the archive's
106
+ `<MediaFilePath>` set and warns on a coarse-identity cross-link
107
+ (`cross_link_warning`) — `linked == total` alone cannot see one.
102
108
  2. Render a probe range; check frame luma at cut boundaries, dissolve
103
109
  midpoints (expect the blend average), and retime windows.
104
110
  3. For audio: RMS per window (silence = -inf is a failed placement).
@@ -12,7 +12,7 @@ that none exists).
12
12
 
13
13
  **Verified on:** DaVinci Resolve Studio 21.0.2
14
14
 
15
- **Totals:** 34 missing capabilities, 44 bugs / unreliable behaviors.
15
+ **Totals:** 35 missing capabilities, 45 bugs / unreliable behaviors.
16
16
 
17
17
  The authoritative source is the runtime-queryable `api_truth` ledger
18
18
  (`resolve_control api_truth "<query>"`); this document is generated from
@@ -221,6 +221,13 @@ equivalent, blocking full automation.
221
221
  - **Workaround / current handling:** To author a non-default start TC offline, patch MediaExtents (drt.assemble spec.startFrame does this) and keep clip Start frames >= the new origin - clips before it are silently dropped on import. For conform, assemble_from_interchange preserveStartTimecode=true anchors at the turnover's real first record frame instead of 01:00:00:00.
222
222
  - **Tags:** timecode, drt, import
223
223
 
224
+ ### Embedded source timecode lives in the clip's MediaStartTime (SECONDS); AAF duplicates audio per channel
225
+
226
+ - **Object:** `Sm2TiVideoClip.MediaStartTime / AAF export`
227
+ - **Behavior:** Two conform-ingest measurements (2026-08-30, rich Resolve 19 AAF export + its sources). (1) A source with embedded timecode is referenced by the timeline clip's <MediaStartTime> in SECONDS (01:00:00:00 -> 3600); a transplant clone keeping the template donor's 0 imports and reads back fine but the render fails with 'Full resolution media not found at 01:00:00:00'. (2) The AAF export carries one event per audio CHANNEL: every A-track event of a dual-mono clip arrives twice with identical ranges.
228
+ - **Workaround / current handling:** capture_media_template harvests mediaStartTime and the native clip elements; drt.assemble clones the source's own captured clip per cut (render-verified: the TC-bearing source plays picture and audio, and the full AAF route renders frame-accurately). Re-capture templates for TC-bearing media. The assemble bridge merges identical audio channel legs (report.audioChannelLegsMerged) instead of refusing them as a same-track overlap.
229
+ - **Tags:** timecode, aaf, audio, drt, silent-failure
230
+
224
231
  ### MediaPool.ImportMedia (current-folder destination only)
225
232
 
226
233
  - **Object:** `MediaPool`
@@ -533,6 +540,13 @@ values, or automation-hostile modal prompts.
533
540
  - **Reference:** [issue #171](https://github.com/samuelgursky/davinci-resolve-mcp/issues/171)
534
541
  - **Tags:** timeline, import, silent-failure, unreliable-return
535
542
 
543
+ ### Timeline import MERGES pool media by a coarse identity - similar files can silently cross-link
544
+
545
+ - **Object:** `ImportTimelineFromFile / media pool`
546
+ - **Behavior:** When a .drt import lands in a project that already holds a media entry whose pool identity blob matches the incoming one, Resolve MERGES them and every clip relinks to the EXISTING media - silently. The identity is coarse: two different files (5.6MB vs 93KB, different names, mtimes 1s apart) had identity blobs byte-identical except internal uuids, and the second file's clips all played the first file's picture (readback showed the wrong clip NAME; measured E33/E34 on 19.1.3.7). Importing the same archive into a FRESH project materialized both files correctly - the merge only bites across imports.
547
+ - **Workaround / current handling:** After importing an authored timeline into a non-empty project, verify per-item file paths (or render probe frames) before trusting the conform - linked==total cannot see a cross-link. Files generated in the same second are the risk class; distinct mtimes distinguish them.
548
+ - **Tags:** import, media-pool, silent-failure, drt
549
+
536
550
  ### Audio tracks cannot be grown in an imported timeline; Fairlight strips live in the pool Sm2Sequence.FieldsBlob
537
551
 
538
552
  - **Object:** `Sm2TiTrack (audio) / FLStudioModelBA`
package/install.py CHANGED
@@ -37,7 +37,7 @@ from src.utils.update_check import (
37
37
 
38
38
  # ─── Version ──────────────────────────────────────────────────────────────────
39
39
 
40
- VERSION = "2.121.0"
40
+ VERSION = "2.123.0"
41
41
  # Only hard floor: mcp[cli] requires Python 3.10+. There is no upper bound —
42
42
  # Resolve's scripting bridge loads into newer interpreters on recent builds
43
43
  # (Python 3.14 verified against Resolve Studio 20.3.2). Older Resolve builds
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "davinci-resolve-mcp",
3
- "version": "2.121.0",
3
+ "version": "2.123.0",
4
4
  "description": "NPM bootstrapper for the DaVinci Resolve MCP Server.",
5
5
  "license": "MIT",
6
6
  "author": "Samuel Gursky <samgursky@gmail.com>",
@@ -78,12 +78,14 @@ async function cutSourceIntoClips(drpInput, opts = {}) {
78
78
  // mirror clones AUDIO even for plain video cuts, so the donor must be
79
79
  // chosen per track type, never per cut flavor (a video element in an
80
80
  // audio track aborts the whole import, measured E31e).
81
- // NATIVE-DONOR PATH (live verification PENDING a stuck Resolve modal
82
- // ended the E31 session before any import of this path could be
83
- // measured cleanly; only caches that carry videoClipElement /
84
- // audioClipElement reach it, so ordinary captures keep the proven donor
85
- // path). Donor clones from splitClipElements come <Element>-wrapped and
86
- // Items concatenation depends on the wrapper.
81
+ // NATIVE-DONOR PATH RENDER-VERIFIED on 19.1.3.7 (E36): a TC-bearing
82
+ // .mov source (embedded 01:00:00:00) renders picture AND audio through a
83
+ // cloned native clip, and the full AAF route (two sources, V2 stacking,
84
+ // merged audio legs) renders every window frame-accurately. Only caches
85
+ // that carry videoClipElement/audioClipElement reach this; ordinary
86
+ // captures keep the proven donor path. Donor clones from
87
+ // splitClipElements come <Element>-wrapped and Items concatenation
88
+ // depends on the wrapper.
87
89
  const native = kind === 'audio' ? cut.donorClipAudio : cut.donorClipVideo;
88
90
  let c = freshDbIds(native ? native.trim() : donor);
89
91
  c = setClipStart(c, cut.startFrame);
@@ -87,7 +87,7 @@ if not logging.getLogger().handlers:
87
87
  handlers=[logging.StreamHandler()],
88
88
  )
89
89
 
90
- VERSION = "2.121.0"
90
+ VERSION = "2.123.0"
91
91
  logger = logging.getLogger("davinci-resolve-mcp")
92
92
  logger.info(f"Starting DaVinci Resolve MCP Server v{VERSION}")
93
93
  logger.info(f"Detected platform: {get_platform()}")
package/src/server.py CHANGED
@@ -11,7 +11,7 @@ Usage:
11
11
  python src/server.py --full # Start the 353-tool granular server instead
12
12
  """
13
13
 
14
- VERSION = "2.121.0"
14
+ VERSION = "2.123.0"
15
15
 
16
16
  import base64
17
17
  import os
@@ -6966,6 +6966,49 @@ def _export_timeline_checked(tl, p: Dict[str, Any]):
6966
6966
  return _run_maybe_background("timeline.export_timeline_checked", p, _work)
6967
6967
 
6968
6968
 
6969
+ def _drt_expected_media_paths(path: str):
6970
+ """Unique <MediaFilePath> values referenced by a .drt/.drp's timeline clips.
6971
+
6972
+ Feeds the cross-link check: Resolve merges pool media by a COARSE identity
6973
+ across imports, so a clip can silently relink to a DIFFERENT pre-existing
6974
+ file (measured: item readback even shows the wrong clip name). The archive
6975
+ is the ground truth for which files the timeline meant.
6976
+ """
6977
+ import zipfile as _zf
6978
+ paths = set()
6979
+ try:
6980
+ with _zf.ZipFile(path) as zf:
6981
+ for n in zf.namelist():
6982
+ if re.search(r"SeqContainer/.+\.xml$", n) or re.search(r"/SeqContainer\d*\.xml$", n):
6983
+ xml = zf.read(n).decode("utf-8", "replace")
6984
+ for m in re.finditer(r"<MediaFilePath>([^<]+)</MediaFilePath>", xml):
6985
+ paths.add(m.group(1))
6986
+ except Exception:
6987
+ return None
6988
+ return paths or None
6989
+
6990
+
6991
+ def _timeline_cross_link_check(tl, expected_paths):
6992
+ """Compare the media files a timeline ACTUALLY links against the expected set."""
6993
+ actual = set()
6994
+ for track_type in ("video", "audio"):
6995
+ try:
6996
+ count = int(tl.GetTrackCount(track_type) or 0)
6997
+ except Exception:
6998
+ count = 0
6999
+ for i in range(1, count + 1):
7000
+ for item in (tl.GetItemListInTrack(track_type, i) or []):
7001
+ try:
7002
+ mpi = item.GetMediaPoolItem()
7003
+ fp = mpi.GetClipProperty("File Path") if mpi else None
7004
+ if fp:
7005
+ actual.add(str(fp))
7006
+ except Exception:
7007
+ pass
7008
+ missing = sorted(x for x in expected_paths if x not in actual)
7009
+ return {"expected": sorted(expected_paths), "actual": sorted(actual), "missing": missing}
7010
+
7011
+
6969
7012
  def _timeline_media_coverage(tl) -> Dict[str, Any]:
6970
7013
  """Count how many timeline items are linked to a Media Pool Item vs. offline.
6971
7014
 
@@ -7427,6 +7470,20 @@ def _import_timeline_checked(proj, mp, p: Dict[str, Any]):
7427
7470
  out["relink"] = relink_result
7428
7471
  if binary_relink_note:
7429
7472
  out["note"] = binary_relink_note
7473
+ if ext in {".drt", ".drp"}:
7474
+ expected_paths = _drt_expected_media_paths(path)
7475
+ if expected_paths:
7476
+ xcheck = _timeline_cross_link_check(imported, expected_paths)
7477
+ if xcheck["missing"]:
7478
+ out["cross_link_warning"] = (
7479
+ "Media files this timeline references are NOT among the files its "
7480
+ f"items actually link to: {', '.join(xcheck['missing'])}. Resolve "
7481
+ "merges pool media by a coarse identity across imports, so clips "
7482
+ "can silently relink to a DIFFERENT pre-existing file (item names "
7483
+ "follow the wrong file too). Import into a fresh project, or "
7484
+ "render-probe before trusting this conform."
7485
+ )
7486
+ out["cross_link_check"] = xcheck
7430
7487
  if media["total"] and media["offline"]:
7431
7488
  msg = f"{media['offline']} of {media['total']} timeline items are offline (no linked media)."
7432
7489
  if is_binary:
@@ -1936,6 +1936,55 @@ API_TRUTH: List[Dict[str, Any]] = [
1936
1936
  "tags": ["timecode", "drt", "import"],
1937
1937
  "submit": "missing",
1938
1938
  },
1939
+ {
1940
+ "symbol": "Timeline import MERGES pool media by a coarse identity - similar files can silently cross-link",
1941
+ "object": "ImportTimelineFromFile / media pool",
1942
+ "reality": "When a .drt import lands in a project that already "
1943
+ "holds a media entry whose pool identity blob matches "
1944
+ "the incoming one, Resolve MERGES them and every clip "
1945
+ "relinks to the EXISTING media - silently. The identity "
1946
+ "is coarse: two different files (5.6MB vs 93KB, "
1947
+ "different names, mtimes 1s apart) had identity blobs "
1948
+ "byte-identical except internal uuids, and the second "
1949
+ "file's clips all played the first file's picture "
1950
+ "(readback showed the wrong clip NAME; measured E33/E34 "
1951
+ "on 19.1.3.7). Importing the same archive into a FRESH "
1952
+ "project materialized both files correctly - the merge "
1953
+ "only bites across imports.",
1954
+ "recommended": "After importing an authored timeline into a "
1955
+ "non-empty project, verify per-item file paths (or "
1956
+ "render probe frames) before trusting the conform - "
1957
+ "linked==total cannot see a cross-link. Files "
1958
+ "generated in the same second are the risk class; "
1959
+ "distinct mtimes distinguish them.",
1960
+ "tags": ["import", "media-pool", "silent-failure", "drt"],
1961
+ "submit": "bug",
1962
+ },
1963
+ {
1964
+ "symbol": "Embedded source timecode lives in the clip's MediaStartTime (SECONDS); AAF duplicates audio per channel",
1965
+ "object": "Sm2TiVideoClip.MediaStartTime / AAF export",
1966
+ "reality": "Two conform-ingest measurements (2026-08-30, rich "
1967
+ "Resolve 19 AAF export + its sources). (1) A source with "
1968
+ "embedded timecode is referenced by the timeline clip's "
1969
+ "<MediaStartTime> in SECONDS (01:00:00:00 -> 3600); a "
1970
+ "transplant clone keeping the template donor's 0 imports "
1971
+ "and reads back fine but the render fails with 'Full "
1972
+ "resolution media not found at 01:00:00:00'. (2) The AAF "
1973
+ "export carries one event per audio CHANNEL: every "
1974
+ "A-track event of a dual-mono clip arrives twice with "
1975
+ "identical ranges.",
1976
+ "recommended": "capture_media_template harvests mediaStartTime and "
1977
+ "the native clip elements; drt.assemble clones the "
1978
+ "source's own captured clip per cut (render-verified: "
1979
+ "the TC-bearing source plays picture and audio, and "
1980
+ "the full AAF route renders frame-accurately). "
1981
+ "Re-capture templates for TC-bearing media. The "
1982
+ "assemble bridge merges identical audio channel legs "
1983
+ "(report.audioChannelLegsMerged) instead of refusing "
1984
+ "them as a same-track overlap.",
1985
+ "tags": ["timecode", "aaf", "audio", "drt", "silent-failure"],
1986
+ "submit": "missing",
1987
+ },
1939
1988
  {
1940
1989
  "symbol": "Audio tracks cannot be grown in an imported timeline; Fairlight strips live in the pool Sm2Sequence.FieldsBlob",
1941
1990
  "object": "Sm2TiTrack (audio) / FLStudioModelBA",