davinci-resolve-mcp 2.104.0 → 2.104.1

package/CHANGELOG.md

@@ -2,6 +2,36 @@
 
 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 v2.104.1
+
+**The job metadata lies too — verify_output now cross-checks the timeline.**
+Finishing v2.104.0's pending live validation exposed a hole in the new
+`render.verify_output`: in the issue #164 case (content before the timeline
+start) Resolve rewrites the render job's own MarkIn/MarkOut down to the
+collapsed extent. Measured live on Studio 19.1.3.7: 96 frames of content
+placed before the start turned an explicit 96-frame mark range into a 1-frame
+job — Complete at 100%, a 1-frame black stub, and a clean duration ratio,
+because the expected duration was computed from the job's own lying range.
+
+The only truthful readback in that state is the timeline items themselves,
+which report their real positions. `verify_output` now finds the job's
+timeline and cross-checks it: video items starting before the timeline's
+start frame warn (the #164 signature by direct evidence), and a mark range
+under half the items' extent warns (the collapse signature). Callers can also
+pass `expected_frames` or `expected_duration_seconds` outright. Live-verified
+both ways on 19.1.3.7: the healthy render verifies clean, the stub now fails
+with both warnings.
+
+### Fixed
+
+- `render.verify_output` no longer trusts the job's MarkIn/MarkOut as the
+  expected duration — the #164 stub previously verified clean at ratio 1.0.
+
+### Changed
+
+- The api_truth recordFrame entry now documents the mark-range rewrite and
+  the item-readback discriminator; the Blackmagic-facing report regenerated.
+
 ## What's New in v2.104.0
 
 The read/write symmetry audit's worklist, worked. PR #162's AST rewrite left

package/README.md

@@ -2,7 +2,7 @@
 
 English | [简体中文](README.zh-CN.md)
 
-[![Version](https://img.shields.io/badge/version-2.104.0-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
+[![Version](https://img.shields.io/badge/version-2.104.1-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-36%20(353%20full)-blue.svg)](#server-modes)

package/README.zh-CN.md

@@ -2,7 +2,7 @@
 
 [English](README.md) | 简体中文
 
-[![Version](https://img.shields.io/badge/version-2.104.0-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
+[![Version](https://img.shields.io/badge/version-2.104.1-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-36%20(353%20full)-blue.svg)](#服务器模式)
@@ -12,7 +12,7 @@
 [![Python](https://img.shields.io/badge/python-3.10+-green.svg)](https://www.python.org/downloads/)
 [![License](https://img.shields.io/badge/license-MIT-blue.svg)](https://opensource.org/licenses/MIT)
 
-> 本翻译对应 v2.104.0 版 README。如与英文原版有出入，以 [英文原版](README.md) 为准。
+> 本翻译对应 v2.104.1 版 README。如与英文原版有出入，以 [英文原版](README.md) 为准。
 
 一个 Model Context Protocol (MCP) 服务器，让 AI 助手通过官方脚本 API 控制 DaVinci Resolve Studio（达芬奇）。它提供完整的 API 覆盖，外加带护栏的工作流助手，涵盖剪辑、媒体池整理、渲染设置、审阅标记、调色、Fusion、Fairlight、项目生命周期任务、扩展开发，以及不碰源媒体的媒体分析。
 

package/install.py

@@ -37,7 +37,7 @@ from src.utils.update_check import (
 
 # ─── Version ──────────────────────────────────────────────────────────────────
 
-VERSION = "2.104.0"
+VERSION = "2.104.1"
 # 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": "2.104.0",
+  "version": "2.104.1",
   "description": "NPM bootstrapper for the DaVinci Resolve MCP Server.",
   "license": "MIT",
   "author": "Samuel Gursky <samgursky@gmail.com>",

package/src/granular/common.py

@@ -87,7 +87,7 @@ if not logging.getLogger().handlers:
         handlers=[logging.StreamHandler()],
     )
 
-VERSION = "2.104.0"
+VERSION = "2.104.1"
 logger = logging.getLogger("davinci-resolve-mcp")
 logger.info(f"Starting DaVinci Resolve MCP Server v{VERSION}")
 logger.info(f"Detected platform: {get_platform()}")

package/src/server.py

@@ -11,7 +11,7 @@ Usage:
     python src/server.py --full       # Start the 353-tool granular server instead
 """
 
-VERSION = "2.104.0"
+VERSION = "2.104.1"
 
 import base64
 import os
@@ -18580,11 +18580,16 @@ def render(action: str, params: Optional[Dict[str, Any]] = None) -> Dict[str, An
       delete_all_jobs() -> {success}
       list_jobs() -> {jobs}
       get_job_status(job_id) -> {status}
-      verify_output(job_id) -> {status, output_path, output_exists, output_size_bytes, output_probe, expected_duration_seconds, duration_ratio, warnings, verified}
+      verify_output(job_id, expected_frames?, expected_duration_seconds?) -> {status, output_path, output_exists, output_size_bytes, output_probe, mark_range_frames, timeline_item_extent_frames, expected_duration_seconds, duration_ratio, warnings, verified}
         JobStatus reports Complete even when the render produced a near-empty
-        stub (content before the timeline start, issue #164) — this checks the
-        actual output file against the job's mark range. Verify BEFORE deleting
-        the job; deleted jobs carry no TargetDir to check.
+        stub (content before the timeline start, issue #164), and Resolve
+        rewrites the job's own MarkIn/MarkOut down to the collapsed extent, so
+        the job metadata lies too. This checks the actual output file, and
+        cross-checks the job's timeline items (which read back their real
+        positions): items before the timeline start and a mark range far
+        smaller than the item extent both warn. Pass expected_frames or
+        expected_duration_seconds when you know what the render should hold.
+        Verify BEFORE deleting the job; deleted jobs carry no TargetDir.
       start(job_ids?, interactive?) -> {success}
       stop() -> {success}
       is_rendering() -> {rendering}
@@ -18675,6 +18680,71 @@ def render(action: str, params: Optional[Dict[str, Any]] = None) -> Dict[str, An
         else:
             warnings.append("Job carries no TargetDir/OutputFilename to verify against.")
         result["output_path"] = output_path
+        # The job's own MarkIn/MarkOut cannot be the only truth source: when
+        # timeline content sits before the timeline's start frame (issue #164)
+        # Resolve rewrites the job's mark range down to the collapsed extent,
+        # so the job metadata lies right along with JobStatus (measured live on
+        # Studio 19.1.3.7: 96 frames of content before start -> job range of 1
+        # frame, Complete at 100%, a 1-frame black stub). Cross-check the
+        # job's timeline items, which DO read back their real positions, and
+        # let the caller override expected duration outright.
+        try:
+            fps = float(str(job.get("FrameRate")))
+        except (TypeError, ValueError):
+            fps = None
+        try:
+            mark_frames = int(job.get("MarkOut")) - int(job.get("MarkIn")) + 1
+        except (TypeError, ValueError):
+            mark_frames = None
+        result["mark_range_frames"] = mark_frames
+        tl_for_job = None
+        try:
+            for i in range(1, int(proj.GetTimelineCount() or 0) + 1):
+                cand = proj.GetTimelineByIndex(i)
+                if cand and cand.GetName() == job.get("TimelineName"):
+                    tl_for_job = cand
+                    break
+        except Exception:
+            tl_for_job = None
+        if tl_for_job is not None:
+            try:
+                tl_start = int(tl_for_job.GetStartFrame())
+                extent_lo, extent_hi, before_start = None, None, 0
+                for ti in range(1, int(tl_for_job.GetTrackCount("video") or 0) + 1):
+                    for it in tl_for_job.GetItemListInTrack("video", ti) or []:
+                        it_start, it_end = int(it.GetStart()), int(it.GetEnd())
+                        extent_lo = it_start if extent_lo is None else min(extent_lo, it_start)
+                        extent_hi = it_end if extent_hi is None else max(extent_hi, it_end)
+                        if it_start < tl_start:
+                            before_start += 1
+                if before_start:
+                    warnings.append(
+                        f"{before_start} video item(s) on timeline "
+                        f"{job.get('TimelineName')!r} start before the timeline's "
+                        f"start frame {tl_start} — the render engine never visits "
+                        "that content (recordFrame counts from absolute frame "
+                        "zero, issue #164). The items read back as placed; the "
+                        "render is a stub."
+                    )
+                if (extent_lo is not None and mark_frames is not None
+                        and mark_frames < (extent_hi - extent_lo) * 0.5):
+                    warnings.append(
+                        f"The job's mark range ({mark_frames} frame(s)) is under "
+                        f"half the timeline's item extent "
+                        f"({extent_hi - extent_lo} frames) — Resolve rewrote the "
+                        "range down to a collapsed extent, which is how a stub "
+                        "render reports a clean duration ratio."
+                    )
+                result["timeline_item_extent_frames"] = (
+                    None if extent_lo is None else extent_hi - extent_lo
+                )
+            except Exception:
+                pass
+        else:
+            warnings.append(
+                f"Timeline {job.get('TimelineName')!r} was not found in the "
+                "project, so item-extent cross-checks were skipped."
+            )
         if output_path:
             exists = os.path.exists(output_path)
             result["output_exists"] = exists
@@ -18684,13 +18754,21 @@ def render(action: str, params: Optional[Dict[str, Any]] = None) -> Dict[str, An
                 if probe:
                     result["output_probe"] = probe
                     duration = probe.get("duration_seconds")
-                    # Expected duration from the job's own mark range and rate.
-                    try:
-                        frames = int(job.get("MarkOut")) - int(job.get("MarkIn")) + 1
-                        fps = float(str(job.get("FrameRate")))
-                        expected = frames / fps if fps > 0 else None
-                    except (TypeError, ValueError):
-                        expected = None
+                    # Expected duration: caller-supplied truth wins; the job's
+                    # mark range is the fallback (see collapse caveat above).
+                    expected = None
+                    if p.get("expected_duration_seconds") is not None:
+                        try:
+                            expected = float(p["expected_duration_seconds"])
+                        except (TypeError, ValueError):
+                            expected = None
+                    elif p.get("expected_frames") is not None and fps:
+                        try:
+                            expected = int(p["expected_frames"]) / fps
+                        except (TypeError, ValueError):
+                            expected = None
+                    elif mark_frames is not None and fps:
+                        expected = mark_frames / fps
                     result["expected_duration_seconds"] = expected
                     if duration is not None and expected and expected > 0:
                         result["duration_ratio"] = round(duration / expected, 4)

package/src/utils/api_truth.py

@@ -1875,10 +1875,20 @@ API_TRUTH: List[Dict[str, Any]] = [
                    "(frame 0 == first frame), while recordFrame and "
                    "TimelineItem.GetStart/GetEnd are absolute. Distinct from "
                    "the null-id entry below, which is a recordFrame landing in "
-                   "an OCCUPIED span rather than before the start. Verified on "
+                   "an OCCUPIED span rather than before the start. Worse, the "
+                   "render JOB's own metadata lies too: Resolve rewrites the "
+                   "job's MarkIn/MarkOut down to the collapsed extent (measured "
+                   "live on 19.1.3.7: 96 frames of content before start -> an "
+                   "explicit 96-frame mark range became a 1-frame job, Complete "
+                   "at 100%, a 1-frame black stub), so output-duration vs "
+                   "job-mark-range checks read clean. The only truthful "
+                   "readback is the timeline items themselves, which report "
+                   "their real GetStart/GetEnd — before the start frame. "
+                   "Verified on "
                    "Studio 20.3.2.9 (v2.17.1 probe: relative record_frame 12 "
-                   "landed at 86400 + 12 = 86412, absolute preserved 86484) and "
-                   "independently on Studio 21.0.4.5 (issue #164).",
+                   "landed at 86400 + 12 = 86412, absolute preserved 86484), "
+                   "independently on Studio 21.0.4.5 (issue #164), and the "
+                   "mark-range rewrite on Studio 19.1.3.7 (2026-08-29).",
         "recommended": "Offset every recordFrame by timeline.GetStartFrame(). "
                        "This server already does it: "
                        "media_pool.append_to_timeline defaults to "
@@ -1888,8 +1898,10 @@ API_TRUTH: List[Dict[str, Any]] = [
                        "value below the timeline start is refused outright. "
                        "When driving the API directly, "
                        "never treat JobStatus Complete as proof a render "
-                       "worked — check the output file's duration, not just "
-                       "that the job finished.",
+                       "worked — and do not trust the job's MarkIn/MarkOut as "
+                       "the expected duration either. render.verify_output "
+                       "(v2.104.1) cross-checks the job against its timeline's "
+                       "item extents and flags items before the start frame.",
         "tags": ["timeline", "edit", "render", "silent-failure", "media-pool"],
         "issue": 164,
     },