davinci-resolve-mcp 2.97.4 → 2.97.6

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.
package/CHANGELOG.md CHANGED
@@ -2,6 +2,96 @@
2
2
 
3
3
  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.
4
4
 
5
+ ## What's New in v2.97.6
6
+
7
+ **On Windows the server could not see the Resolve it was driving.** Reported in
8
+ [#150](https://github.com/samuelgursky/davinci-resolve-mcp/issues/150) with the
9
+ root cause traced, the fix proposed, and a case table — all of it correct.
10
+
11
+ ### Fixed
12
+
13
+ - **`runtime_mode` reported `running: false, instances: 0` on every stock
14
+ Windows install.** WMIC wraps a command line in double quotes when the
15
+ executable path contains spaces, which the default install path always does
16
+ (`"C:\Program Files\Blackmagic Design\DaVinci Resolve\Resolve.exe"`). The
17
+ line therefore *ends* with `"`, and `_is_resolve_command()` required it to end
18
+ with `Resolve.exe`. The trailing-flag stripper above the test did not help: it
19
+ removes ` -flag` tokens and leaves the closing quote as the final character.
20
+ A leading quote is now read for what it is — the executable is what sits
21
+ inside the first quoted span, and everything after the closing quote is
22
+ arguments. That also tightens rejection: `"…\cmd.exe" /c start … Resolve.exe`
23
+ is a launcher, not an instance.
24
+
25
+ The reading was the visible half. The consequential half is that
26
+ `get_resolve()` asks this same question before auto-launching, precisely so a
27
+ failed connect to a live Resolve does not open a second application — the
28
+ guard whose comment records being "reported three times before it was
29
+ traced". On Windows its input was a permanent `False`, so the path it exists
30
+ to block was open: connect fails (modal dialog, mid-launch, scripting toggled
31
+ off) → "nothing is running" → launch a second Resolve. `_not_connected_error()`
32
+ reads the same signal, so a Windows user whose Resolve *was* running got the
33
+ "not running, auto-launch failed, check your Studio install" text instead of
34
+ the preference or bridge fix that actually applied. `headless` was
35
+ unreachable too — it is only computed once something is found running, so
36
+ `-nogui` instances were indistinguishable on Windows.
37
+
38
+ - **A successful install ended in a traceback under a redirected stdout.** The
39
+ installer prints box-drawing and check-mark glyphs. A Windows console carries
40
+ them, but redirecting stdout falls back to the locale code page — cp1252 by
41
+ default — and `print(f" {'─' * 50}")` in the summary raised
42
+ `UnicodeEncodeError` after every client had already been configured, so a
43
+ working run looked like a failed one
44
+ (`npx davinci-resolve-mcp setup --clients manual 2>&1 | tail`). Streams that
45
+ cannot encode those glyphs are now reconfigured to UTF-8 with
46
+ `errors="replace"`; a console that can already carry them is left alone
47
+ rather than re-encoded underneath the user.
48
+
49
+ ### Added
50
+
51
+ - `tests.test_headless_runtime` covers the quoted Windows command line —
52
+ bare, trailing-whitespace (what WMIC actually prints), and `-nogui` — plus
53
+ the quoted-launcher line that must still be rejected, so the fix cannot
54
+ become a substring test by another route. `tests.test_cdl_and_install_config`
55
+ gains `ConsoleEncodingTests`, including a child interpreter run with
56
+ `PYTHONIOENCODING=cp1252` and a piped stdout: the exact shape of the reported
57
+ failure. Both suites fail on the pre-fix code.
58
+
59
+ ## What's New in v2.97.5
60
+
61
+ **`npm ci` was failing outright, and nothing in the release path noticed.**
62
+ No server behavior changed; this is packaging and release-process hardening.
63
+
64
+ ### Fixed
65
+
66
+ - **`package-lock.json` was seven releases stale.** It still carried
67
+ `2.90.0`, and — the part that actually broke things — two
68
+ `optionalDependencies` added since then, `js-yaml` and `pg`, were never
69
+ locked. `npm ci` refuses to install at all when the lockfile and
70
+ `package.json` disagree, so every reproducible install path failed with
71
+ `EUSAGE — Missing: js-yaml@4.3.1 from lock file` (plus `pg` and its nine
72
+ transitive deps). CI, fresh contributor clones, and container builds all hit
73
+ it. `npm install` and `npm publish` resolve independently of the lockfile
74
+ and stayed green throughout, which is why it survived seven releases.
75
+ Regenerated with `npm install --package-lock-only`; `npm ci` now installs
76
+ 175 packages clean.
77
+
78
+ ### Added
79
+
80
+ - **`tests.test_import::test_package_lock_in_sync`** — asserts both version
81
+ fields in the lockfile match `package.json`, and that the root
82
+ `dependencies` / `devDependencies` / `optionalDependencies` blocks match
83
+ exactly. The second half is the one that matters: the version fields being
84
+ right is not evidence `npm ci` works, and dependency drift is what actually
85
+ breaks it. Verified to fail on each drift mode independently.
86
+
87
+ ### Changed
88
+
89
+ - `docs/process/release-process.md` lists `package-lock.json` under "Files To
90
+ Update" with the regeneration command, and Required Validation now
91
+ regenerates the lockfile before `test_import` reads it. The lockfile was
92
+ never on the checklist, which is why the drift was never a step anyone
93
+ skipped — it was a step that did not exist.
94
+
5
95
  ## What's New in v2.97.4
6
96
 
7
97
  **Drift is caught at the edit, not at publish time.** Tooling and docs only —
package/README.md CHANGED
@@ -2,7 +2,7 @@
2
2
 
3
3
  English | [简体中文](README.zh-CN.md)
4
4
 
5
- [![Version](https://img.shields.io/badge/version-2.97.4-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
5
+ [![Version](https://img.shields.io/badge/version-2.97.6-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
6
6
  [![npm](https://img.shields.io/npm/v/davinci-resolve-mcp.svg?label=npm&color=CB3837)](https://www.npmjs.com/package/davinci-resolve-mcp)
7
7
  [![API Coverage](https://img.shields.io/badge/API%20Coverage-100%25-brightgreen.svg)](docs/reference/api-coverage.md)
8
8
  [![Tools](https://img.shields.io/badge/MCP%20Tools-35%20(353%20full)-blue.svg)](#server-modes)
package/README.zh-CN.md CHANGED
@@ -2,7 +2,7 @@
2
2
 
3
3
  [English](README.md) | 简体中文
4
4
 
5
- [![Version](https://img.shields.io/badge/version-2.97.4-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
5
+ [![Version](https://img.shields.io/badge/version-2.97.6-blue.svg)](https://github.com/samuelgursky/davinci-resolve-mcp/releases)
6
6
  [![npm](https://img.shields.io/npm/v/davinci-resolve-mcp.svg?label=npm&color=CB3837)](https://www.npmjs.com/package/davinci-resolve-mcp)
7
7
  [![API Coverage](https://img.shields.io/badge/API%20Coverage-100%25-brightgreen.svg)](docs/reference/api-coverage.md)
8
8
  [![Tools](https://img.shields.io/badge/MCP%20Tools-35%20(353%20full)-blue.svg)](#服务器模式)
@@ -12,7 +12,7 @@
12
12
  [![Python](https://img.shields.io/badge/python-3.10+-green.svg)](https://www.python.org/downloads/)
13
13
  [![License](https://img.shields.io/badge/license-MIT-blue.svg)](https://opensource.org/licenses/MIT)
14
14
 
15
- > 本翻译对应 v2.97.4 版 README。如与英文原版有出入,以 [英文原版](README.md) 为准。
15
+ > 本翻译对应 v2.97.6 版 README。如与英文原版有出入,以 [英文原版](README.md) 为准。
16
16
 
17
17
  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。
18
18
 
@@ -26,6 +26,14 @@ Every release bump must update all version surfaces:
26
26
  - `src/granular/common.py`
27
27
  - `install.py`
28
28
  - `package.json`
29
+ - `package-lock.json` — regenerate with `npm install --package-lock-only`, do
30
+ not hand-edit the version. It carries the version in two places and, more
31
+ importantly, the resolved dependency tree. When it disagrees with
32
+ `package.json`, **`npm ci` refuses to install at all** (`EUSAGE — Missing: X
33
+ from lock file`), which breaks CI, fresh clones, and container builds while
34
+ `npm install` and `npm publish` stay green and hide it. The
35
+ `tests.test_import` guard `test_package_lock_in_sync` fails the suite on
36
+ either drift.
29
37
  - README version badge
30
38
  - README current stats or latest-release summary when they changed
31
39
  - `README.zh-CN.md` — the Simplified Chinese translation. Update its version
@@ -62,6 +70,7 @@ Always run static checks before release:
62
70
 
63
71
  ```bash
64
72
  venv/bin/python tests/test_import.py
73
+ npm install --package-lock-only --no-audit --no-fund # re-stage package-lock.json if it moved
65
74
  venv/bin/python scripts/audit_api_parity.py
66
75
  venv/bin/python scripts/gen_api_limitations.py --check
67
76
  node scripts/agent-rules/generate.mjs --check
@@ -72,6 +81,13 @@ npm pack --dry-run
72
81
  git diff --check
73
82
  ```
74
83
 
84
+ `test_import` carries `test_package_lock_in_sync`, which is the actual gate on the
85
+ lockfile: it asserts both version fields and the root dependency blocks match
86
+ `package.json`. Run the `npm install --package-lock-only` line above first so the
87
+ regeneration is in the working tree when the test reads it — that ordering is why
88
+ the check is a regeneration followed by a test, not a `git diff --exit-code`,
89
+ which would fire on the release bump's own legitimate change.
90
+
75
91
  `test_duplicate_definitions` asserts no module-level name is defined twice under
76
92
  `src/`. A second `def foo` silently replaces the first, and in a module the size
77
93
  of `src/server.py` the two can be thousands of lines apart with different
package/install.py CHANGED
@@ -37,7 +37,7 @@ from src.utils.update_check import (
37
37
 
38
38
  # ─── Version ──────────────────────────────────────────────────────────────────
39
39
 
40
- VERSION = "2.97.4"
40
+ VERSION = "2.97.6"
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
@@ -46,6 +46,42 @@ VERSION = "2.97.4"
46
46
  SUPPORTED_PYTHON_MIN = (3, 10)
47
47
  PYTHON_ABI_RISK_MIN = (3, 13)
48
48
 
49
+ # ─── Console encoding ─────────────────────────────────────────────────────────
50
+ # This installer prints box-drawing and check-mark glyphs. On Windows a console
51
+ # stdout carries them fine, but a *redirected* stdout falls back to the locale
52
+ # code page — cp1252 on a default Windows install — and the first '─' raises
53
+ # UnicodeEncodeError. It fires at the summary, after every client is already
54
+ # configured, so a successful install ends in a traceback and reads as a failed
55
+ # run. Reported in #150 against
56
+ # `npx davinci-resolve-mcp setup --clients manual 2>&1 | tail`.
57
+ #
58
+ # Only streams that cannot already carry the glyphs are touched, so a correctly
59
+ # configured console keeps its own encoding. 'replace' is belt-and-braces: no
60
+ # output path is worth a traceback.
61
+
62
+ _GLYPH_PROBE = "─→✓⊘•"
63
+
64
+ def _ensure_glyph_capable_stdio():
65
+ for stream in (sys.stdout, sys.stderr):
66
+ reconfigure = getattr(stream, "reconfigure", None)
67
+ encoding = getattr(stream, "encoding", None)
68
+ if reconfigure is None or not encoding:
69
+ continue
70
+ try:
71
+ _GLYPH_PROBE.encode(encoding)
72
+ continue
73
+ except (LookupError, UnicodeEncodeError):
74
+ pass
75
+ try:
76
+ reconfigure(encoding="utf-8", errors="replace")
77
+ except Exception:
78
+ try:
79
+ reconfigure(errors="replace")
80
+ except Exception:
81
+ pass
82
+
83
+ _ensure_glyph_capable_stdio()
84
+
49
85
  # ─── Colors (disabled on Windows cmd without ANSI support) ────────────────────
50
86
 
51
87
  def _supports_color():
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "davinci-resolve-mcp",
3
- "version": "2.97.4",
3
+ "version": "2.97.6",
4
4
  "description": "NPM bootstrapper for the DaVinci Resolve MCP Server.",
5
5
  "license": "MIT",
6
6
  "author": "Samuel Gursky <samgursky@gmail.com>",
@@ -87,7 +87,7 @@ if not logging.getLogger().handlers:
87
87
  handlers=[logging.StreamHandler()],
88
88
  )
89
89
 
90
- VERSION = "2.97.4"
90
+ VERSION = "2.97.6"
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 = "2.97.4"
14
+ VERSION = "2.97.6"
15
15
 
16
16
  import base64
17
17
  import os
@@ -83,8 +83,13 @@ def _process_lines() -> Optional[List[str]]:
83
83
  return None
84
84
 
85
85
 
86
+ def _matches_pattern(executable: str) -> bool:
87
+ """Does this bare executable path name a Resolve application?"""
88
+ return any(executable.endswith(pattern) for pattern in RESOLVE_PROCESS_PATTERNS)
89
+
90
+
86
91
  def _is_resolve_command(line: str) -> bool:
87
- """Is this command line a Resolve *executable*, not merely a mention of one?
92
+ r"""Is this command line a Resolve *executable*, not merely a mention of one?
88
93
 
89
94
  A plain substring test matches any process whose command line happens to
90
95
  contain the path — including a shell running a script that references it.
@@ -96,8 +101,22 @@ def _is_resolve_command(line: str) -> bool:
96
101
  flags. So strip trailing flag tokens and require what remains to *end* with
97
102
  the pattern. That survives the spaces in "DaVinci Resolve.app" (no splitting
98
103
  on whitespace) while rejecting a path buried mid-command.
104
+
105
+ Windows quotes that path. WMIC prints the executable wrapped in double
106
+ quotes whenever it contains spaces, which the default install path always
107
+ does (`"C:\Program Files\Blackmagic Design\DaVinci Resolve\Resolve.exe"`),
108
+ so the line ends in `"` and `endswith("Resolve.exe")` was false on every
109
+ stock Windows machine — `runtime_mode` reported nothing running while the
110
+ same server was driving that very instance, and the second-instance guard
111
+ in `get_resolve()` lost its input. Reported in #150. A leading quote means
112
+ the executable is exactly what sits inside the first quoted span; anything
113
+ after the closing quote is arguments, and the flag loop never sees it.
99
114
  """
100
115
  text = line.strip()
116
+ if text.startswith('"'):
117
+ close = text.find('"', 1)
118
+ if close > 1:
119
+ return _matches_pattern(text[1:close])
101
120
  while True:
102
121
  stripped = text.rstrip()
103
122
  cut = stripped.rfind(" -")
@@ -109,7 +128,7 @@ def _is_resolve_command(line: str) -> bool:
109
128
  if not candidate:
110
129
  break
111
130
  text = candidate
112
- return any(text.endswith(pattern) for pattern in RESOLVE_PROCESS_PATTERNS)
131
+ return _matches_pattern(text)
113
132
 
114
133
 
115
134
  def resolve_processes() -> Optional[List[str]]: