davinci-resolve-mcp 4.10.7 → 4.11.1

package/CHANGELOG.md

@@ -2,6 +2,64 @@
 
 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 v4.11.1 — `.cube` input ranges, `"false"` import flags and the caption character cap
+
+Three fixes from @Dev-next-gen (#286, #287, #288), each with a reproduction and a
+regression test.
+
+### Fixes
+
+- **A 3D `.cube` carrying `LUT_3D_INPUT_RANGE` could not be read** (#288).
+  `read_cube` knew `DOMAIN_MIN`/`DOMAIN_MAX` but not Resolve's own header form,
+  so the line fell through to the data-row branch and the whole file was refused
+  as a "non-numeric entry". `read_lut_summary` and `attenuate_lut_file` both
+  failed on it. The keyword is now read as one domain for all three channels;
+  attenuation still refuses a non-unit range, exactly as it does for
+  `DOMAIN_MIN`/`DOMAIN_MAX`.
+  - **On landing:** a short or non-numeric `LUT_3D_INPUT_RANGE` escaped as a bare
+    `ValueError`, and a short `DOMAIN_MIN`/`DOMAIN_MAX` produced a silently short
+    domain list. All three headers now go through one reader that raises
+    `CubeLutError` naming the file, line and keyword.
+- **`timeline(action="import_timeline_checked")` read `"false"` as true** (#287).
+  `sanitize_media` / `relink_media` and `import_source_clips` were read with bare
+  `bool()`, so `sanitize_media="false"` still handed Resolve the sanitized copy
+  instead of the file passed, and `import_source_clips="false"` reached
+  `ImportTimelineFromFile` as `importSourceClips: True`. Both now go through
+  `_coerce_bool`, as the rest of `server.py` has since v4.8.20.
+- **Caption orphan rebalancing could break `max_chars_per_line`** (#286).
+  Moving the last word of the penultimate line onto a one-word final line never
+  re-checked the cap, so "we had a wonderful conversation" at 20 characters came
+  out with a 22-character second line in the SRT/WebVTT. The move now happens
+  only when the merged line fits; a lone last word is the lesser fault.
+
+## What's New in v4.11.0 — `rename_plan` collides names a case-insensitive volume treats as one
+
+### Fixes
+
+- **`media(action="rename_plan")` missed collisions that only differ in case
+  or Unicode form.** macOS (default APFS) and Windows volumes are
+  case-insensitive, and APFS also ignores Unicode normalization, so renaming
+  `take_A.mov` to `A.mov` overwrites an existing `a.mov`. The plan compared
+  names byte-for-byte and reported that rename as free (`collisions: 0`).
+  - Targets are now compared after NFC normalization and lowercasing, so
+    `A.mov` / `a.mov` and `café` spelled precomposed or with a combining
+    accent each count as one file.
+  - A case-only rename of a single file (`clip.MOV` → `clip.mov`) is still a
+    `rename`, not a collision.
+
+### New parameter
+
+- `rename_plan` takes `caseSensitive` (default `false`). Pass `true` only for
+  a case-sensitive volume to restore exact comparison. The result reports the
+  mode used as `caseSensitive`.
+
+### Validation
+
+- Offline planning only; nothing calls Resolve or touches files. New
+  `media-ops` test fails on the old code and passes now; `group-grade-read`
+  gains a case for a group name with `"`, `<` and `>`, as Studio 19.1.3.7
+  escapes them.
+
 ## What's New in v4.10.7 — `retargetDRT` moves the `<ProjectVersion>` import gate, not only the comment stamp
 
 ### Fixes

package/README.md

@@ -2,7 +2,7 @@
 
 English | [简体中文](README.zh-CN.md)
 
-[![Version](https://img.shields.io/badge/version-4.10.7-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
+[![Version](https://img.shields.io/badge/version-4.11.1-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-37%20(389%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-4.10.7-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
+[![Version](https://img.shields.io/badge/version-4.11.1-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-37%20(389%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 v4.10.7 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 v4.11.1 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 = "4.10.7"
+VERSION = "4.11.1"
 # 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": "4.10.7",
+  "version": "4.11.1",
   "description": "NPM bootstrapper for the DaVinci Resolve MCP Server.",
   "license": "MIT",
   "author": "Samuel Gursky <samgursky@gmail.com>",

package/resolve-advanced/server/media-ops.mjs

@@ -121,7 +121,9 @@ const CAMERA_ORIGINAL_RE = /^([A-Z]\d{3}[_ ]?[CR]\d{3,4}|[A-Z]\d{3}_\d{8}_C\d+|M
 /**
  * Dry-run rename plan (regex find/replace). REFUSES camera originals (sidecar/DB only).
  * @param {string[]} names basenames to rename
- * @param {{find:string, replace:string, allowCameraOriginals?:boolean, cameraOriginalRe?:RegExp}} rule
+ * @param {{find:string, replace:string, allowCameraOriginals?:boolean, cameraOriginalRe?:RegExp, caseSensitive?:boolean}} rule
+ *   caseSensitive defaults to false: macOS (APFS) and Windows volumes treat names that
+ *   differ only in case or Unicode normalization as one file, so those collide.
  */
 export function renamePlan(names, rule = {}) {
   let re;
@@ -133,27 +135,30 @@ export function renamePlan(names, rule = {}) {
   const camRe = rule.cameraOriginalRe || CAMERA_ORIGINAL_RE;
   const plan = [];
   const targets = new Map();
+  const caseSensitive = rule.caseSensitive === true;
+  const key = (n) => (caseSensitive ? n : n.normalize('NFC').toLowerCase());
   // A name the plan leaves in place still occupies it — a refused camera original
   // above all — so it counts as a target too, or a rename onto it reads as free.
   for (const name of names) {
     const isCam = camRe.test(name);
     if (isCam && !rule.allowCameraOriginals) {
       plan.push({ from: name, action: 'refuse-camera-original', note: 'camera original — rename via sidecar/DB only, never the file' });
-      targets.set(name, (targets.get(name) || 0) + 1);
+      targets.set(key(name), (targets.get(key(name)) || 0) + 1);
       continue;
     }
     const to = name.replace(re, rule.replace ?? '');
-    targets.set(to, (targets.get(to) || 0) + 1);
+    targets.set(key(to), (targets.get(key(to)) || 0) + 1);
     plan.push({ from: name, to, action: to === name ? 'noop' : 'rename' });
   }
   // Flag collisions (two sources → one target).
-  for (const p of plan) if (p.action === 'rename' && targets.get(p.to) > 1) p.action = 'collision';
+  for (const p of plan) if (p.action === 'rename' && targets.get(key(p.to)) > 1) p.action = 'collision';
   const collisions = plan.filter((p) => p.action === 'collision');
   return {
     plan,
     renameCount: plan.filter((p) => p.action === 'rename').length,
     refused: plan.filter((p) => p.action === 'refuse-camera-original').length,
     collisions: collisions.length,
+    caseSensitive,
     dryRun: true,
   };
 }

package/resolve-advanced/server/tools/media.mjs

@@ -47,6 +47,10 @@ const renameSchema = z.object({
   find: z.string(),
   replace: z.string().optional(),
   allowCameraOriginals: z.boolean().optional(),
+  caseSensitive: z
+    .boolean()
+    .optional()
+    .describe('Default false: names differing only in case/Unicode form collide, as on macOS and Windows volumes. Pass true only for a case-sensitive volume.'),
 });
 
 const reelSchema = z.object({ reels: z.array(z.string()), pad: z.number().optional() });
@@ -83,7 +87,7 @@ export const mediaTool = {
     }
     if (action === 'rename_plan') {
       const p = renameSchema.parse(args);
-      return renamePlan(p.names, { find: p.find, replace: p.replace, allowCameraOriginals: p.allowCameraOriginals });
+      return renamePlan(p.names, { find: p.find, replace: p.replace, allowCameraOriginals: p.allowCameraOriginals, caseSensitive: p.caseSensitive });
     }
     if (action === 'reel_normalize') {
       const p = reelSchema.parse(args);

package/src/granular/common.py

@@ -91,7 +91,7 @@ if not logging.getLogger().handlers:
         handlers=[logging.StreamHandler()],
     )
 
-VERSION = "4.10.7"
+VERSION = "4.11.1"
 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 377-tool granular server instead
 """
 
-VERSION = "4.10.7"
+VERSION = "4.11.1"
 
 import base64
 import os
@@ -7699,7 +7699,7 @@ def _import_timeline_checked(proj, mp, p: Dict[str, Any]):
     # abort Resolve's scripting-API import and leave the timeline fully offline.
     # The original is only read; the sanitized copy lands in a temp dir, so the
     # temp-path guard never applies to it.
-    sanitize_requested = bool(p.get("sanitize_media") or p.get("relink_media"))
+    sanitize_requested = _coerce_bool(p.get("sanitize_media")) or _coerce_bool(p.get("relink_media"))
     binary_relink_note = None
     _binary_has_roots = bool(p.get("relink_search_roots") or p.get("search_roots"))
     if is_binary and _binary_has_roots:
@@ -7730,7 +7730,7 @@ def _import_timeline_checked(proj, mp, p: Dict[str, Any]):
     if p.get("timeline_name") and "timelineName" not in options:
         options["timelineName"] = p["timeline_name"]
     if "import_source_clips" in p and "importSourceClips" not in options:
-        options["importSourceClips"] = bool(p["import_source_clips"])
+        options["importSourceClips"] = _coerce_bool(p["import_source_clips"])
 
     # Optional fuzzy relink: when search roots are given, missing-at-original-path
     # clips are matched against on-disk media (exact/ext-agnostic/normalized/reel)

package/src/utils/captions.py

@@ -83,13 +83,19 @@ def _wrap(text: str, max_chars: int, max_lines: int) -> Optional[List[str]]:
     return lines if len(lines) <= max_lines else None
 
 
-def _no_orphan(lines: List[str]) -> List[str]:
-    """Rebalance so no line is a lone word when a neighbour could give one up."""
+def _no_orphan(lines: List[str], max_chars: int) -> List[str]:
+    """Rebalance so no line is a lone word when a neighbour could give one up.
+
+    It cannot when the merged line would break the character cap `_wrap` has
+    just enforced; an orphan is the lesser of the two faults.
+    """
     if len(lines) < 2:
         return lines
     if len(lines[-1].split()) == 1 and len(lines[-2].split()) > 2:
         head = lines[-2].split()
-        lines = lines[:-2] + [" ".join(head[:-1]), f"{head[-1]} {lines[-1]}"]
+        merged = f"{head[-1]} {lines[-1]}"
+        if len(merged) <= max_chars:
+            lines = lines[:-2] + [" ".join(head[:-1]), merged]
     return lines
 
 
@@ -147,7 +153,7 @@ def build_blocks(
                 _word_time(current[-1], "end_seconds", "end")
                 or _word_time(current[-1], "start_seconds", "start")
             ),
-            "lines": _no_orphan(lines),
+            "lines": _no_orphan(lines, max_chars_per_line),
             "word_count": len(current),
         })
         current.clear()

package/src/utils/cube_lut.py

@@ -28,7 +28,7 @@ from __future__ import annotations
 
 import os
 import re
-from typing import Any, Dict, Optional, Tuple
+from typing import Any, Dict, List, Optional, Tuple
 
 try:
     import numpy as _np
@@ -56,6 +56,19 @@ def _require_numpy() -> None:
         raise CubeLutError("numpy is required for LUT arithmetic (pip install numpy)")
 
 
+def _header_floats(path: str, line_number: int, line: str, count: int) -> List[float]:
+    """The `count` numbers after a header keyword, or a CubeLutError naming the line."""
+    values = line.split()[1:count + 1]
+    try:
+        parsed = [float(value) for value in values]
+    except ValueError:
+        parsed = []
+    if len(parsed) != count:
+        keyword = line.split()[0]
+        raise CubeLutError(f"{path}:{line_number}: malformed {keyword} (expected {count} numbers)")
+    return parsed
+
+
 def read_cube(path: str) -> Dict[str, Any]:
     """Parse a 3D `.cube` file into `{size, table, domain_min, domain_max, title}`.
 
@@ -100,10 +113,15 @@ def read_cube(path: str) -> Dict[str, Any]:
                 title = match.group(1) if match else ""
                 continue
             if upper.startswith("DOMAIN_MIN"):
-                domain_min = [float(value) for value in line.split()[1:4]]
+                domain_min = _header_floats(path, line_number, line, 3)
                 continue
             if upper.startswith("DOMAIN_MAX"):
-                domain_max = [float(value) for value in line.split()[1:4]]
+                domain_max = _header_floats(path, line_number, line, 3)
+                continue
+            if upper.startswith("LUT_3D_INPUT_RANGE"):
+                # Resolve's form of DOMAIN_MIN/MAX: one min and one max for all channels.
+                low, high = _header_floats(path, line_number, line, 2)
+                domain_min, domain_max = [low] * 3, [high] * 3
                 continue
             parts = line.split()
             if len(parts) != 3: