davinci-resolve-mcp 2.104.8 → 2.104.10

package/CHANGELOG.md

@@ -2,6 +2,59 @@
 
 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 v2.104.10
+
+Stones turned on the live-validation backlog, on Studio 19.1.3.7.
+
+**AAF live import, validated at last** (marked "NOT live-validated" since
+2026-07-06): an EXPORT_AAF/EXPORT_AAF_NEW round trip imports cleanly with
+importSourceClips=false, lands fully offline (the documented turnover shape),
+and preserves the start timecode. The naming matrix across import formats is
+now complete and in api_truth — FCP7 XML ignores timelineName (internal name
+wins, the #171 trap); AAF honours timelineName when given, else its internal
+name; OTIO honours timelineName; .drt names the timeline after the file. Only
+FCP7 exhibits the returned-existing trap. The AAF post-import relink leg was
+exercised too: under importSourceClips=false it correctly reports "no Media
+Pool Items to relink" — the API relinks pool items, and none exist on that
+path — so its precondition is now stated instead of assumed.
+
+**safe_quick_export verifies its output.** RenderWithQuickExport's status
+dict was the last render surface trusted without a file check: a success
+status that wrote nothing read as an export. The files that actually landed
+in TargetDir are now listed with size and ffprobe duration, and a success
+status with no new file flips to an error.
+
+**Housekeeping:** the delete-locked scratch project from the v2.104.7 session
+is gone — the pre-restart DeleteProject had returned False while actually
+succeeding, the documented DeleteProject lie caught in the wild. A clean
+Resolve quit/relaunch verified the wedge entry's other half: Quit() works
+when no orphaned render holds the pipeline.
+
+## What's New in v2.104.9
+
+**The NTSC coverage gap, closed.** The conform fixtures were integer-rate
+only — which is how parseEDL ran exact-rate timecode math against a
+nominal-rate writer for years (fixed in v2.104.6, convention measured against
+Resolve's own GetStartFrame). New fixtures now exercise the pipeline at
+29.97: a broadcast-start EDL parses to nominal frames (an NDF minute is 1800
+frames, butt cuts stay gapless), the EDL write→parse round trip is
+frame-identical, media-inventory's tc↔frames round trip is the identity at
+all three NTSC rates, and drop-frame pins to the canonical values
+(01:00:00;00 → 107892 — the number that haunted the #168 saga, now living
+where it belongs). Cross-language pin tests assert the Python converters
+(_timecode_to_frame_id, multicam) and the Node converters agree on the same
+canonical values, so a change that moves one side fails the other side's
+suite.
+
+**Python dependency stack: audited to zero.** pip-audit over the dev venv
+found and cleared advisories in urllib3, requests, python-multipart,
+setuptools, starlette (0.52 → 1.6 — the MCP SDK tolerated the major, full
+suite green), pyjwt, pydantic-settings, pygments, pillow, idna, msgpack,
+cryptography, pip itself, and torch 2.13 (with the matching torchvision).
+Also found: the venv's mcp SDK was at 1.27.0, BELOW the repo's own >=1.29
+floor, and carrying a CVE — now 1.29.1. Both stacks (npm and pip) report
+zero known vulnerabilities, with clean resolver constraints.
+
 ## What's New in v2.104.8
 
 More laps: the aggregation class swept to completion, and the dependency

package/README.md

@@ -2,7 +2,7 @@
 
 English | [简体中文](README.zh-CN.md)
 
-[![Version](https://img.shields.io/badge/version-2.104.8-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
+[![Version](https://img.shields.io/badge/version-2.104.10-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-2.104.8-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
+[![Version](https://img.shields.io/badge/version-2.104.10-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 v2.104.8 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 v2.104.10 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/docs/reference/api-limitations.md

@@ -521,7 +521,7 @@ values, or automation-hostile modal prompts.
 
 - **Object:** `MediaPool`
 - **Signature:** `(filePath, {timelineName, importSourceClips, ...}) -> Timeline`
-- **Behavior:** For FCP7 XML, the sequence name INSIDE the file wins over the timelineName import option. When the internal name matches an existing timeline, the call returns that EXISTING timeline — no error, no new timeline — so an export→edit→re-import loop keying uniqueness on the option 'succeeds' while operating on one timeline forever (issue #171, Studio 21.0.4.5; wrapper behavior verified on 19.1.3.7). Distinct from the documented repeated-timelineName None return: here the option is fresh and the file's name is stale.
+- **Behavior:** For FCP7 XML, the sequence name INSIDE the file wins over the timelineName import option. When the internal name matches an existing timeline, the call returns that EXISTING timeline — no error, no new timeline — so an export→edit→re-import loop keying uniqueness on the option 'succeeds' while operating on one timeline forever (issue #171, Studio 21.0.4.5; wrapper behavior verified on 19.1.3.7). Distinct from the documented repeated-timelineName None return: here the option is fresh and the file's name is stale. THE NAMING AUTHORITY DIFFERS PER FORMAT (all measured on 19.1.3.7): FCP7 XML ignores timelineName entirely (internal <name> wins); AAF honours timelineName when given and falls back to its internal name; OTIO honours timelineName; and .drt names the timeline after the FILE (see the .drt entry below). Only FCP7 exhibits the returned-existing trap.
 - **Workaround / current handling:** Rewrite the <sequence><name> inside the file to the intended name before importing — timeline.import_timeline_checked does this automatically for FCP7 XML and errors when a non-rewritable format still returns an existing timeline. Never treat a truthy return as proof of creation; check the returned timeline's id against the pre-import set.
 - **Reference:** [issue #171](https://github.com/samuelgursky/davinci-resolve-mcp/issues/171)
 - **Tags:** timeline, import, silent-failure, unreliable-return

package/install.py

@@ -37,7 +37,7 @@ from src.utils.update_check import (
 
 # ─── Version ──────────────────────────────────────────────────────────────────
 
-VERSION = "2.104.8"
+VERSION = "2.104.10"
 # 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": "2.104.8",
+  "version": "2.104.10",
   "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 = "2.104.8"
+VERSION = "2.104.10"
 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 = "2.104.8"
+VERSION = "2.104.10"
 
 import base64
 import os
@@ -18975,8 +18975,40 @@ def _safe_quick_export(proj, p: Dict[str, Any]):
         return {"success": False, "validation": validation}
     if p.get("dry_run") or not p.get("allow_render", False):
         return _ok(would_render=False, preset=preset, params=params, validation=validation)
+    before = set()
+    if target_dir and os.path.isdir(target_dir):
+        before = set(os.listdir(target_dir))
     status = _ser(proj.RenderWithQuickExport(preset, params))
-    return {"success": not (isinstance(status, dict) and status.get("error")), "status": status, "params": params}
+    out: Dict[str, Any] = {
+        "success": not (isinstance(status, dict) and status.get("error")),
+        "status": status,
+        "params": params,
+    }
+    # RenderWithQuickExport's status dict is trusted nowhere else in this
+    # repo's render surface anymore (issue #164 taught that lesson) — check
+    # what actually landed on disk. Best-effort: TargetDir is known, so a
+    # successful export must have produced at least one new file there.
+    if out["success"] and target_dir and os.path.isdir(target_dir):
+        new_files = sorted(set(os.listdir(target_dir)) - before)
+        outputs = []
+        for name in new_files:
+            path = os.path.join(target_dir, name)
+            if not os.path.isfile(path):
+                continue
+            rec: Dict[str, Any] = {"path": path, "size_bytes": os.path.getsize(path)}
+            probe = _ffprobe_media_summary(path)
+            if probe:
+                rec["duration_seconds"] = probe.get("duration_seconds")
+            outputs.append(rec)
+        out["outputs"] = outputs
+        if not outputs:
+            out["success"] = False
+            out["error"] = (
+                "RenderWithQuickExport reported success but wrote no file to "
+                f"{target_dir} — treat the status dict as unreliable and check "
+                "the Deliver page's render queue."
+            )
+    return out
 
 
 def _export_render_boundary_report(proj, p: Dict[str, Any]):
@@ -19049,7 +19081,10 @@ def render(action: str, params: Optional[Dict[str, Any]] = None) -> Dict[str, An
         way to read back — see the SetRenderSettings api_truth entry.
       render_job_lifecycle_probe(target_dir, settings?, format?, codec?, custom_name?) -> {success, job_id, status_before_delete}
       quick_export_capabilities() -> {presets, safe_params, guards}
-      safe_quick_export(preset, target_dir?|params?, custom_name?, dry_run?, allow_render?) -> {success, status}
+      safe_quick_export(preset, target_dir?|params?, custom_name?, dry_run?, allow_render?) -> {success, status, outputs}
+        After a live export, the files that actually landed in TargetDir are
+        listed with size and ffprobe duration; a success status that wrote no
+        file flips success to false (the issue #164 trust lesson).
       export_render_boundary_report(include_matrix?, max_pairs?, include_quick_export?) -> {capabilities, settings, matrix?}
       list_delivery_targets(tier?, check_availability?) -> {targets, tiers, schema_version}
       resolve_delivery_target(target, overrides?) -> {format_id, codec_id, settings, qc_spec}

package/src/utils/api_truth.py

@@ -1866,7 +1866,13 @@ API_TRUTH: List[Dict[str, Any]] = [
                    "(issue #171, Studio 21.0.4.5; wrapper behavior verified on "
                    "19.1.3.7). Distinct from the documented repeated-"
                    "timelineName None return: here the option is fresh and the "
-                   "file's name is stale.",
+                   "file's name is stale. THE NAMING AUTHORITY DIFFERS PER "
+                   "FORMAT (all measured on 19.1.3.7): FCP7 XML ignores "
+                   "timelineName entirely (internal <name> wins); AAF honours "
+                   "timelineName when given and falls back to its internal "
+                   "name; OTIO honours timelineName; and .drt names the "
+                   "timeline after the FILE (see the .drt entry below). Only "
+                   "FCP7 exhibits the returned-existing trap.",
         "recommended": "Rewrite the <sequence><name> inside the file to the "
                        "intended name before importing — "
                        "timeline.import_timeline_checked does this "