claude-token-saver 3.28.0 → 3.28.2

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/README.en.md CHANGED
@@ -397,12 +397,18 @@ Three situations, three different interception points:
397
397
  That second row is measured, not assumed: a `.pdf` Read fires the hook, and a `.pptx` Read in the same session leaves no hook log entry at all.
398
398
 
399
399
  ```bash
400
- claude-token-saver doc2md install-converter # markitdown + editing libs + the .fig parser
401
- claude-token-saver doc2md on # register the Read hook
400
+ claude-token-saver doc2md on # register the hooks (the converter installs itself)
402
401
  claude-token-saver doc2md # check converter + hook registration
403
402
  claude-token-saver doc2md report.pptx # convert by hand and see the result
403
+ claude-token-saver doc2md install-converter # only to get the install out of the way early
404
404
  ```
405
405
 
406
+ **The converter installs itself.** Any rollout step a person has to be told about is a step some of them skip, so the converter installs in the background the moment a document first shows up, and converts as soon as it is ready. Measured: about 30s for the first document (15s install plus markitdown's first import), then 3.7s for a new document and 0.1s on a cache hit. The `.fig` parser installs in half a second on the first Figma file.
407
+
408
+ It installs on first use rather than at `install` time: the venv is 47MB, and someone who never opens a document should not pay for it. Set `CTS_DOC2MD_NO_AUTOINSTALL=1` to turn the automatic install off.
409
+
410
+ **Python 3.10+ is required** — markitdown's own floor, and macOS still ships 3.9 as `/usr/bin/python3`. The venv is built on an interpreter chosen by version rather than by PATH order. Built on 3.9, pip resolves markitdown to a 2019 placeholder release (0.0.1a1): the install looks like it worked and every conversion then dies at import. This was found by walking into it. When nothing on the machine is new enough, the message points at `brew install python` instead of at an install command that cannot succeed.
411
+
406
412
  The converter goes into a venv this tool owns (`<state dir>/doc2md-venv`): no system interpreter is touched, and uninstalling the CLI takes it along. An existing markitdown on `uv tool` or `PATH` is preferred over building a new one.
407
413
 
408
414
  Conversion is [markitdown](https://github.com/microsoft/markitdown). Slide numbers, heading levels, tables, speaker notes and per-sheet headings all survive, and non-Latin text comes through intact.
@@ -455,6 +461,23 @@ Conversion is one-way — editing the cached `.md` changes nothing in the source
455
461
 
456
462
  All four formats were exercised end to end on 2026-09-06: 10 docx run replacements plus three consecutive re-saves, a pptx bar-to-line chart swap with an added data point, xlsx value edits and a new row, and a fig text edit with re-encode and re-parse. In every case the original was byte-identical afterwards and the re-converted copy showed the change. One caveat: removing a chart shape from a pptx leaves the old chart XML part orphaned — PowerPoint ignores it, but delete the part and its rels for a clean file. Charts and images never appear in a conversion, so visual edits must be confirmed in the application itself.
457
463
 
464
+ ### Locked documents, and Windows
465
+
466
+ **A password-protected document is a state, not an error.** Office encrypts by wrapping the package in an OLE compound file rather than a zip, so opening one as a zip used to report "not a zip file" — which reads as a broken download and sends the user after the wrong problem. It is now identified before conversion:
467
+
468
+ ```
469
+ ✗ encrypted: password-protected Office file (OLE-wrapped)
470
+ ✗ encrypted: password-protected PDF
471
+ ```
472
+
473
+ The model is told to ask for an unlocked copy. This tool never asks for or stores a password, and never blocks the original `Read`, so work continues either way. A PDF that merely restricts printing still opens and still converts — checked against a false positive — and a legacy `.xls`, which is an OLE file by design, is not mistaken for an encrypted one.
474
+
475
+ **Windows is supported.** For teams with Windows machines:
476
+
477
+ - The Python search uses the `py -3` launcher. `python3` is rarely on PATH there, and a bare `python` may be the Store alias stub that opens a web page instead of running anything. Venv interpreters are looked for at `Scripts\python.exe`.
478
+ - The `.fig` parser installs through `npm.cmd` via the shell, and the package spec dropped its caret (`openfig-core@0.4.x`): in cmd.exe `^` is the escape character and never reaches npm.
479
+ - The background install and every child process set `windowsHide`, so no console window appears in the middle of someone's prompt.
480
+
458
481
  `claude-token-saver doc2md --clean` empties the conversion cache; `doc2md off` removes the hook. Removal filters for this tool's own entry, so anything else you registered under `PreToolUse` stays.
459
482
 
460
483
  ## 🌐 Behind a gateway (Bedrock / Vertex)
package/README.md CHANGED
@@ -389,12 +389,18 @@ doc2md 는 그 파일을 한 번 변환해 두고 원본 대신 변환본을 읽
389
389
  두 번째 줄의 제약은 실측으로 확인한 것입니다. `.pdf` 를 Read 하면 훅이 실행되고, 같은 세션에서 `.pptx` 를 Read 하면 훅 로그에 아무 기록도 남지 않습니다.
390
390
 
391
391
  ```bash
392
- claude-token-saver doc2md install-converter # markitdown·편집 라이브러리·.fig 파서 설치
393
- claude-token-saver doc2md on # Read 훅 등록
392
+ claude-token-saver doc2md on # 등록 (변환기는 첫 문서에서 자동 설치)
394
393
  claude-token-saver doc2md # 변환기·훅 등록 상태 확인
395
394
  claude-token-saver doc2md 보고서.pptx # 직접 변환해 결과 확인
395
+ claude-token-saver doc2md install-converter # 설치를 미리 끝내 두고 싶을 때만
396
396
  ```
397
397
 
398
+ **변환기는 알아서 깔립니다.** 팀에 배포할 때 각자 설치 명령을 실행하게 만들면 그 단계에서 빠지는 사람이 생깁니다. 그래서 문서가 처음 등장하는 시점에 변환기가 백그라운드로 설치되고, 설치가 끝나는 대로 곧바로 변환합니다. 실측으로 첫 문서는 약 30초(설치 15초 + markitdown 최초 임포트), 이후로는 새 문서 3.7초, 캐시 적중 0.1초입니다. `.fig` 파서는 첫 Figma 파일에서 0.5초 만에 깔립니다.
399
+
400
+ 설치는 `install` 단계가 아니라 첫 사용 시점에 합니다. venv 가 47MB 라서, 문서를 다루지 않는 사람은 낼 이유가 없는 비용입니다. 자동 설치를 끄려면 `CTS_DOC2MD_NO_AUTOINSTALL=1` 을 설정하십시오.
401
+
402
+ **파이썬 3.10 이상이 필요합니다.** markitdown 의 요구 사항이고, macOS 기본 `/usr/bin/python3` 는 3.9 입니다. 이 도구는 PATH 순서를 따르지 않고 3.10 이상인 인터프리터를 골라 venv 를 만듭니다. 3.9 로 만들면 pip 가 markitdown 을 2019 년 자리표시자 릴리스(0.0.1a1)로 해석해서, 설치는 성공한 것처럼 보이지만 모든 변환이 임포트 단계에서 죽습니다. 실제로 이 함정을 밟고 잡았습니다. 3.10 이상이 아예 없으면 설치 명령을 안내하는 대신 `brew install python` 을 안내합니다.
403
+
398
404
  변환기는 도구 전용 venv(`<상태 디렉터리>/doc2md-venv`)에 설치합니다. 시스템 파이썬을 건드리지 않고, CLI를 지우면 함께 사라집니다. 이미 `uv tool` 이나 다른 경로에 markitdown 이 있으면 그쪽을 먼저 씁니다.
399
405
 
400
406
  변환은 [markitdown](https://github.com/microsoft/markitdown)이 담당하며, 슬라이드 번호와 제목 계층, 표, 발표자 노트, 시트 구분이 모두 남습니다. 한글도 깨지지 않습니다.
@@ -451,6 +457,23 @@ PDF 는 첨부하면 모델이 내용을 그대로 읽습니다. 대신 페이
451
457
 
452
458
  네 형식 모두 실제로 몇 바퀴 돌려 검증했습니다(2026-09-06): docx 텍스트 치환 10건과 3회 연속 재저장, pptx 막대→꺾은선 차트 교체와 데이터 행 추가, xlsx 값 정정·행 추가, fig 텍스트 수정·재인코드·재파싱. 전 케이스에서 원본은 바이트 그대로였고, 수정본 재변환에 변경 내용이 반영됐습니다. 한 가지 주의: pptx 에서 차트 도형을 제거하면 옛 차트 XML 파트가 고아로 남습니다. PowerPoint 는 무시하지만, 깔끔히 하려면 파트와 rels 도 지우십시오. 차트·이미지 같은 시각 요소는 변환본에 잡히지 않으므로, 시각 편집의 최종 확인은 해당 프로그램에서 해야 합니다.
453
459
 
460
+ ### 암호가 걸린 문서와 Windows
461
+
462
+ **암호 문서는 오류가 아니라 상태로 다룹니다.** 사내에서 받는 문서 중 일부는 암호가 걸려 있습니다. 이전에는 암호 걸린 docx 를 "손상된 zip"이라고 알려서 사용자가 원인을 엉뚱한 곳에서 찾게 만들었습니다. 암호가 걸린 Office 문서는 zip 이 아니라 OLE 복합 문서로 저장되기 때문입니다. 지금은 변환 전에 판별해서 이렇게 알립니다.
463
+
464
+ ```
465
+ ✗ encrypted: password-protected Office file (OLE-wrapped)
466
+ ✗ encrypted: password-protected PDF
467
+ ```
468
+
469
+ 모델에게는 암호를 푼 사본을 사용자에게 요청하라고 안내합니다. 이 도구는 암호를 묻지도 저장하지도 않습니다. 어느 경우에도 원본 `Read` 를 막지 않으므로 작업이 중단되지 않습니다. 열람은 자유롭고 인쇄만 제한된 PDF 는 암호 문서가 아니므로 그대로 변환합니다(오탐 확인 완료). 구형 `.xls` 도 원래 OLE 형식이라 암호로 오인하지 않습니다.
470
+
471
+ **Windows 를 지원합니다.** 사내에 Windows 사용자가 있어 다음을 맞췄습니다.
472
+
473
+ - 파이썬 탐색이 `py -3` 런처를 씁니다. Windows 에서는 `python3` 가 PATH 에 없는 경우가 많고, 맨 `python` 은 실행 대신 마이크로소프트 스토어를 여는 별칭 스텁일 수 있습니다. venv 기반 인터프리터도 `Scripts\python.exe` 경로로 찾습니다.
474
+ - `.fig` 파서 설치가 `npm.cmd` 를 셸로 호출합니다. 그리고 패키지 지정자에서 캐럿을 뺐습니다(`openfig-core@0.4.x`). cmd.exe 에서 `^` 는 이스케이프 문자라 npm 에 닿기 전에 먹힙니다.
475
+ - 백그라운드 자동 설치와 모든 하위 프로세스에 `windowsHide` 를 걸어, 프롬프트 도중에 콘솔 창이 튀어나오지 않게 했습니다.
476
+
454
477
  `claude-token-saver doc2md --clean` 으로 변환 캐시를 비우고, `doc2md off` 로 훅을 제거합니다. 훅 해제는 자기 항목만 골라 지우므로 `PreToolUse` 에 등록해 둔 다른 훅은 그대로 남습니다.
455
478
 
456
479
  ## 🌐 Bedrock·Vertex 경유 환경
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "claude-token-saver",
3
- "version": "3.28.0",
3
+ "version": "3.28.2",
4
4
  "description": "Route the easy work your expensive Claude model keeps repeating down to haiku/sonnet — post-hoc session analysis, no realtime router, no extra LLM calls.",
5
5
  "type": "module",
6
6
  "bin": {
@@ -43,6 +43,55 @@ def fail(reason, detail=""):
43
43
  sys.exit(0)
44
44
 
45
45
 
46
+ # A password-protected OOXML file is not a zip at all: Office wraps the whole
47
+ # package in an OLE compound file whose streams hold the ciphertext. Opening it
48
+ # as a zip therefore reports "not a zip file", which reads as a broken download
49
+ # and sends the user looking for the wrong problem.
50
+ OLE_MAGIC = b"\xd0\xcf\x11\xe0\xa1\xb1\x1a\xe1"
51
+
52
+
53
+ def encryption_problem(path, ext):
54
+ """('encrypted', detail) when the file is password-protected, else None."""
55
+ try:
56
+ with open(path, "rb") as fh:
57
+ head = fh.read(8)
58
+ except OSError:
59
+ return None
60
+
61
+ # Legacy .xls is an OLE file by design, so the magic alone proves nothing
62
+ # there. For the modern formats it can only mean encryption.
63
+ if ext in ZIP_EXTS and head == OLE_MAGIC:
64
+ return ("encrypted", "password-protected Office file (OLE-wrapped)")
65
+
66
+ # Some producers keep the zip container and put the ciphertext inside it.
67
+ if ext in ZIP_EXTS:
68
+ try:
69
+ import zipfile
70
+ with zipfile.ZipFile(path) as z:
71
+ names = z.namelist()
72
+ if any(n.startswith("EncryptedPackage") for n in names):
73
+ return ("encrypted", "password-protected Office file")
74
+ except Exception:
75
+ return None
76
+
77
+ if ext == ".pdf":
78
+ try:
79
+ from pdfminer.pdfparser import PDFParser
80
+ from pdfminer.pdfdocument import PDFDocument
81
+ with open(path, "rb") as fh:
82
+ doc = PDFDocument(PDFParser(fh))
83
+ # An empty owner password is the ordinary "printing restricted"
84
+ # case, which extracts fine. Only a document that refuses to open
85
+ # counts as encrypted here.
86
+ if doc.encryption is not None and not doc.is_extractable:
87
+ return ("encrypted", "password-protected PDF")
88
+ except Exception as e:
89
+ if "password" in str(type(e).__name__).lower() or "password" in str(e).lower():
90
+ return ("encrypted", "password-protected PDF")
91
+ return None
92
+ return None
93
+
94
+
46
95
  def check_zip(path):
47
96
  """Classify an archive before opening it as a document.
48
97
 
@@ -189,6 +238,13 @@ def main():
189
238
  fail("missing", path)
190
239
 
191
240
  ext = os.path.splitext(path)[1].lower()
241
+
242
+ # Checked before anything else opens the file: an encrypted document is a
243
+ # normal thing to receive, not a failure to report as corruption.
244
+ locked = encryption_problem(path, ext)
245
+ if locked:
246
+ fail(locked[0], locked[1])
247
+
192
248
  if ext in ZIP_EXTS:
193
249
  problem = check_zip(path)
194
250
  if problem:
@@ -224,6 +280,11 @@ def main():
224
280
  result = MarkItDown().convert(path)
225
281
  text = (result.text_content or "").strip()
226
282
  except Exception as e:
283
+ # markitdown surfaces the password failure from whichever backend hit
284
+ # it, so the type name is the reliable part.
285
+ blob = (type(e).__name__ + " " + str(e)).lower()
286
+ if "password" in blob or "encrypted" in blob:
287
+ fail("encrypted", "the file is password-protected")
227
288
  fail("convert-failed", e)
228
289
 
229
290
  if not text:
package/src/doc2md.cjs CHANGED
@@ -26,7 +26,7 @@ const fs = require('node:fs');
26
26
  const os = require('node:os');
27
27
  const path = require('node:path');
28
28
  const crypto = require('node:crypto');
29
- const { spawnSync } = require('node:child_process');
29
+ const { spawn, spawnSync } = require('node:child_process');
30
30
  const ledger = require('./doc2md-ledger.cjs');
31
31
 
32
32
  // Formats where the original is of no use to the model. Images are absent
@@ -226,28 +226,62 @@ function writeCache(filePath, markdown, extra) {
226
226
  * memoized for the life of this process, and the caller memoizes across
227
227
  * processes through the notice file.
228
228
  */
229
+ // The readiness test for an interpreter. Importing the *class* matters: a pip
230
+ // install creates the package directory long before it finishes writing into
231
+ // it, so a bare `import markitdown` reports success mid-install and the
232
+ // conversion then fails with "cannot import name 'MarkItDown'". Measured
233
+ // during the first-use auto-install, where the window is about a second wide.
234
+ const PROBE = 'from markitdown import MarkItDown';
235
+
236
+ const IS_WINDOWS = process.platform === 'win32';
237
+
238
+ /**
239
+ * Interpreters to try, as `{ bin, args }` so the Windows launcher can carry
240
+ * its version flag. `py -3` is how a Windows box usually reaches Python at
241
+ * all: `python3` is rarely on PATH there, and a bare `python` may be the App
242
+ * Execution Alias stub that opens the Microsoft Store instead of running
243
+ * anything.
244
+ */
245
+ function interpreterCandidates() {
246
+ const out = [];
247
+ if (process.env.CTS_DOC2MD_PYTHON) out.push({ bin: process.env.CTS_DOC2MD_PYTHON, args: [] });
248
+ // The tool's own venv, created by `doc2md install-converter`. First because
249
+ // it is the only one this tool controls: telling people to `pip install`
250
+ // into the system interpreter is how a token-saving CLI ends up owning a
251
+ // break in someone else's project.
252
+ out.push({ bin: managedPython(), args: [] });
253
+ if (IS_WINDOWS) {
254
+ out.push(
255
+ { bin: path.join(os.homedir(), '.local', 'share', 'uv', 'tools', 'markitdown', 'Scripts', 'python.exe'), args: [] },
256
+ { bin: 'py', args: ['-3'] },
257
+ { bin: 'python', args: [] },
258
+ );
259
+ } else {
260
+ out.push(
261
+ { bin: path.join(os.homedir(), '.local', 'share', 'uv', 'tools', 'markitdown', 'bin', 'python'), args: [] },
262
+ { bin: path.join(os.homedir(), '.local', 'bin', 'markitdown-python'), args: [] },
263
+ { bin: 'python3', args: [] },
264
+ { bin: 'python', args: [] },
265
+ );
266
+ }
267
+ return out;
268
+ }
269
+
229
270
  let interpreterCache;
230
271
  function findInterpreter() {
231
272
  if (interpreterCache !== undefined) return interpreterCache;
232
- const candidates = [];
233
- if (process.env.CTS_DOC2MD_PYTHON) candidates.push(process.env.CTS_DOC2MD_PYTHON);
234
- candidates.push(
235
- // The tool's own venv, created by `doc2md install-converter`. First
236
- // because it is the only one this tool controls: telling people to
237
- // `pip install` into the system interpreter is how a token-saving CLI
238
- // ends up owning a break in someone else's project.
239
- managedPython(),
240
- path.join(os.homedir(), '.local', 'share', 'uv', 'tools', 'markitdown', 'bin', 'python'),
241
- path.join(os.homedir(), '.local', 'bin', 'markitdown-python'),
242
- 'python3',
243
- 'python',
244
- );
245
- for (const bin of candidates) {
273
+ for (const { bin, args } of interpreterCandidates()) {
246
274
  try {
247
- const probe = spawnSync(bin, ['-c', 'import markitdown'], { timeout: 20_000, stdio: 'ignore' });
275
+ const probe = spawnSync(bin, [...args, '-c', PROBE], {
276
+ timeout: 20_000,
277
+ stdio: 'ignore',
278
+ windowsHide: true,
279
+ });
248
280
  if (probe.status === 0) {
249
- interpreterCache = bin;
250
- return bin;
281
+ // Only a bare interpreter is cached as a string; the launcher form
282
+ // keeps its flag, since dropping it would run the wrong Python.
283
+ interpreterCache = args.length ? { bin, args } : bin;
284
+ return interpreterCache;
251
285
  }
252
286
  } catch { /* candidate unusable, try the next */ }
253
287
  }
@@ -255,6 +289,11 @@ function findInterpreter() {
255
289
  return null;
256
290
  }
257
291
 
292
+ /** Split a `findInterpreter()` result into the pair spawnSync wants. */
293
+ function interpreterParts(found) {
294
+ return typeof found === 'string' ? { bin: found, args: [] } : found;
295
+ }
296
+
258
297
  const CONVERTER = path.join(__dirname, '..', 'presets', 'doc2md', 'convert.py');
259
298
 
260
299
  /** Path to the interpreter inside the venv this tool manages. */
@@ -281,19 +320,72 @@ const EDIT_LIBS = ['python-pptx', 'python-docx', 'openpyxl'];
281
320
  * for, it goes somewhere this tool owns, so uninstalling the CLI takes the
282
321
  * whole thing with it and no system interpreter is touched.
283
322
  */
323
+ /**
324
+ * A Python new enough to run markitdown (3.10+), or null.
325
+ *
326
+ * Explicit version names are tried before the bare `python3` so a modern
327
+ * Homebrew interpreter wins over the system one regardless of PATH order.
328
+ */
329
+ function findVenvBase() {
330
+ const candidates = IS_WINDOWS
331
+ ? [
332
+ // The launcher first, asked for each version in turn: it knows about
333
+ // installs that never touched PATH, which is the normal case on
334
+ // Windows.
335
+ { bin: 'py', args: ['-3.14'] }, { bin: 'py', args: ['-3.13'] },
336
+ { bin: 'py', args: ['-3.12'] }, { bin: 'py', args: ['-3.11'] },
337
+ { bin: 'py', args: ['-3.10'] }, { bin: 'py', args: ['-3'] },
338
+ { bin: 'python', args: [] },
339
+ ]
340
+ : [
341
+ { bin: 'python3.14', args: [] }, { bin: 'python3.13', args: [] },
342
+ { bin: 'python3.12', args: [] }, { bin: 'python3.11', args: [] },
343
+ { bin: 'python3.10', args: [] },
344
+ { bin: 'python3', args: [] }, { bin: 'python', args: [] },
345
+ ];
346
+ for (const { bin, args } of candidates) {
347
+ const r = spawnSync(bin, [...args, '-c', 'import sys; print("%d.%d" % sys.version_info[:2])'], {
348
+ encoding: 'utf8',
349
+ timeout: 20_000,
350
+ windowsHide: true,
351
+ });
352
+ if (r.status !== 0) continue;
353
+ const [major, minor] = String(r.stdout).trim().split('.').map(Number);
354
+ if (major > 3 || (major === 3 && minor >= 10)) {
355
+ return { bin, args, version: `python ${major}.${minor}` };
356
+ }
357
+ }
358
+ return null;
359
+ }
360
+
284
361
  function installConverter({ onProgress = () => {} } = {}) {
285
362
  const venv = path.join(userDataDir(), 'doc2md-venv');
286
363
  const target = managedPython();
287
364
 
288
365
  if (!fs.existsSync(target)) {
289
- onProgress(`creating ${venv}`);
290
- let created = false;
291
- for (const base of ['python3', 'python']) {
292
- const r = spawnSync(base, ['-m', 'venv', venv], { encoding: 'utf8', timeout: 180_000 });
293
- if (r.status === 0) { created = true; break; }
366
+ // The base interpreter is chosen by version, not by whichever `python3`
367
+ // comes first on PATH. markitdown needs 3.10+, and macOS still ships 3.9
368
+ // as /usr/bin/python3: building the venv on that one installs a
369
+ // seven-year-old placeholder release (0.0.1a1) that has no MarkItDown
370
+ // class in it, and every conversion then fails at import time. Measured
371
+ // on this machine, where /usr/bin/python3 precedes Homebrew's 3.14.
372
+ const base = findVenvBase();
373
+ if (!base) {
374
+ return {
375
+ ok: false,
376
+ reason: 'no-python',
377
+ detail: 'markitdown needs Python 3.10 or newer; none was found on PATH '
378
+ + '(macOS /usr/bin/python3 is 3.9 — install a newer one, e.g. `brew install python`)',
379
+ };
294
380
  }
295
- if (!created) {
296
- return { ok: false, reason: 'no-python', detail: 'no python3 with the venv module on PATH' };
381
+ onProgress(`creating ${venv} (${base.version})`);
382
+ const r = spawnSync(base.bin, [...(base.args || []), '-m', 'venv', venv], {
383
+ encoding: 'utf8',
384
+ timeout: 180_000,
385
+ windowsHide: true,
386
+ });
387
+ if (r.status !== 0) {
388
+ return { ok: false, reason: 'no-python', detail: (r.stderr || 'venv creation failed').slice(0, 300) };
297
389
  }
298
390
  }
299
391
 
@@ -301,6 +393,7 @@ function installConverter({ onProgress = () => {} } = {}) {
301
393
  const install = spawnSync(target, ['-m', 'pip', 'install', '--quiet', MARKITDOWN_SPEC, ...EDIT_LIBS], {
302
394
  encoding: 'utf8',
303
395
  timeout: 900_000,
396
+ windowsHide: true,
304
397
  });
305
398
  if (install.status !== 0) {
306
399
  return { ok: false, reason: 'pip-failed', detail: (install.stderr || '').slice(0, 400) };
@@ -308,7 +401,7 @@ function installConverter({ onProgress = () => {} } = {}) {
308
401
 
309
402
  // The probe is the actual acceptance test: pip can exit 0 and still leave an
310
403
  // interpreter that cannot import what was asked for.
311
- const probe = spawnSync(target, ['-c', 'import markitdown'], { timeout: 60_000, stdio: 'ignore' });
404
+ const probe = spawnSync(target, ['-c', PROBE], { timeout: 60_000, stdio: 'ignore', windowsHide: true });
312
405
  if (probe.status !== 0) {
313
406
  return { ok: false, reason: 'import-failed', detail: 'installed, but markitdown does not import' };
314
407
  }
@@ -320,8 +413,8 @@ function installConverter({ onProgress = () => {} } = {}) {
320
413
  /**
321
414
  * Convert one file. Returns `{ ok: true, cacheFile, meta }`, or
322
415
  * `{ ok: false, reason, detail }` where reason is one of:
323
- * no-markitdown | too-large | sensitive | unsafe-archive | no-text |
324
- * convert-failed | timeout
416
+ * no-markitdown | python-too-old | no-figparser | encrypted | too-large |
417
+ * sensitive | unsafe-archive | no-text | convert-failed | timeout
325
418
  *
326
419
  * Every failure is a reason to leave the original Read alone, never to break
327
420
  * it. That is the whole contract with the hook.
@@ -370,7 +463,14 @@ function convert(filePath, { converter = CONVERTER, python: pythonOverride = nul
370
463
  // Figma files take the Node converter; everything else goes to markitdown.
371
464
  if (path.extname(filePath).toLowerCase() === '.fig') {
372
465
  const fig2md = require('./fig2md.cjs');
373
- const result = spawnFigConvert(fig2md, filePath);
466
+ let result = spawnFigConvert(fig2md, filePath);
467
+ // The .fig parser is an npm install of a few hundred KB, fast enough to
468
+ // wait for inline the first time a Figma file turns up.
469
+ if (!result.ok && result.reason === 'no-figparser'
470
+ && process.env.CTS_DOC2MD_NO_AUTOINSTALL !== '1') {
471
+ fig2md.installFigParser(userDataDir());
472
+ result = spawnFigConvert(fig2md, filePath);
473
+ }
374
474
  if (!result.ok) return result;
375
475
  const written = writeCache(filePath, result.markdown, {
376
476
  note: result.note, truncated: false, rows: 0, pages: 0, markupBytes: 0,
@@ -381,13 +481,29 @@ function convert(filePath, { converter = CONVERTER, python: pythonOverride = nul
381
481
  // The override exists so tests can drive a stub converter with any Python at
382
482
  // all: the normal search insists the interpreter can import markitdown,
383
483
  // which would make the whole path untestable without the real package.
384
- const python = pythonOverride || findInterpreter();
484
+ let python = pythonOverride || findInterpreter();
485
+ if (!python && !pythonOverride) {
486
+ // Telling someone to run the install command is wrong when the install
487
+ // cannot succeed on this machine. A too-old interpreter is a different
488
+ // problem with a different fix, so it gets its own reason rather than
489
+ // hiding behind "no converter".
490
+ if (!findVenvBase()) {
491
+ return {
492
+ ok: false,
493
+ reason: 'python-too-old',
494
+ detail: 'markitdown needs Python 3.10 or newer; none was found on PATH',
495
+ };
496
+ }
497
+ if (ensureConverterInstalled()) python = findInterpreter();
498
+ }
385
499
  if (!python) return { ok: false, reason: 'no-markitdown' };
386
500
 
387
- const run = spawnSync(python, [converter, filePath], {
501
+ const py = interpreterParts(python);
502
+ const run = spawnSync(py.bin, [...py.args, converter, filePath], {
388
503
  encoding: 'utf8',
389
504
  timeout: CONVERT_TIMEOUT_MS,
390
505
  maxBuffer: MAX_MARKDOWN_BYTES * 4,
506
+ windowsHide: true,
391
507
  });
392
508
  if (run.error && run.error.code === 'ETIMEDOUT') return { ok: false, reason: 'timeout' };
393
509
  if (run.status !== 0) {
@@ -421,6 +537,75 @@ function convert(filePath, { converter = CONVERTER, python: pythonOverride = nul
421
537
  return { ok: true, cached: false, cacheFile: written.cacheFile, meta: written.meta };
422
538
  }
423
539
 
540
+ /**
541
+ * First-use install, so nobody has to be told to run a setup command.
542
+ *
543
+ * A team rollout dies on any step a person has to be told about, so the
544
+ * converter installs itself the first time a document actually shows up. It
545
+ * is still lazy rather than part of `install`: the venv is 47MB, and someone
546
+ * who never opens a document should never pay for it.
547
+ *
548
+ * The install runs detached and the caller waits only briefly. A cold install
549
+ * measured 6 seconds on a fast connection, but a corporate network can be far
550
+ * slower, and a hook that blocks a prompt for a minute is worse than a
551
+ * document that converts on the next turn. So: start it, wait up to
552
+ * `waitMs`, and if it is still going, say so and let this turn proceed
553
+ * without the conversion.
554
+ *
555
+ * Returns true when a converter is ready to use right now.
556
+ */
557
+ function ensureConverterInstalled({ waitMs = 15_000 } = {}) {
558
+ if (process.env.CTS_DOC2MD_NO_AUTOINSTALL === '1') return false;
559
+ if (findInterpreter()) return true;
560
+
561
+ const lock = path.join(userDataDir(), 'doc2md-install.lock');
562
+ let running = false;
563
+ try {
564
+ const started = JSON.parse(fs.readFileSync(lock, 'utf8')).startedAt;
565
+ // A lock older than the pip timeout is a crashed run, not a live one.
566
+ running = Number.isFinite(started) && Date.now() - started < 900_000;
567
+ } catch { /* no lock, or an unreadable one: treat as not running */ }
568
+
569
+ if (!running) {
570
+ try {
571
+ fs.mkdirSync(userDataDir(), { recursive: true });
572
+ fs.writeFileSync(lock, JSON.stringify({ startedAt: Date.now() }));
573
+ const cli = path.join(__dirname, '..', 'bin', 'cli.js');
574
+ // Detached, so a session that ends mid-install does not take the
575
+ // install with it — the next session finds it finished.
576
+ const child = spawn(process.execPath, [cli, 'doc2md', 'install-converter'], {
577
+ detached: true,
578
+ stdio: 'ignore',
579
+ // Without this Windows pops a console window for the install, in the
580
+ // middle of someone's prompt.
581
+ windowsHide: true,
582
+ });
583
+ child.unref();
584
+ } catch {
585
+ try { fs.rmSync(lock, { force: true }); } catch { /* best effort */ }
586
+ return false;
587
+ }
588
+ }
589
+
590
+ // Poll cheaply: the interpreter file appearing is the first sign, and the
591
+ // import probe is the acceptance test. interpreterCache has to be cleared
592
+ // or the memoized null from the top of this function would stick.
593
+ const deadline = Date.now() + waitMs;
594
+ while (Date.now() < deadline) {
595
+ if (fs.existsSync(managedPython())) {
596
+ interpreterCache = undefined;
597
+ if (findInterpreter()) {
598
+ try { fs.rmSync(lock, { force: true }); } catch { /* best effort */ }
599
+ return true;
600
+ }
601
+ }
602
+ // A synchronous sleep, because every caller here is synchronous. 400ms
603
+ // keeps the poll count low over a 15s wait.
604
+ Atomics.wait(new Int32Array(new SharedArrayBuffer(4)), 0, 0, 400);
605
+ }
606
+ return false;
607
+ }
608
+
424
609
  /** Where the "markitdown is not installed" notice records that it was shown. */
425
610
  function noticePath() {
426
611
  return path.join(userDataDir(), 'doc2md-notice.json');
@@ -490,9 +675,26 @@ function decideForRead(context, opts = {}) {
490
675
  markNoticeShown();
491
676
  return {
492
677
  deny: false,
493
- reason: `[doc2md] ${name} 를 변환하려 했으나 markitdown 설치되어 있지 않습니다.\n`
494
- + ` 설치: ${INSTALL_HINT}\n`
495
- + ' 설치 전까지는 원본을 그대로 읽습니다. 안내는 번만 표시됩니다.',
678
+ reason: `[doc2md] ${name} 를 변환할 변환기를 지금 설치하고 있습니다(첫 실행에만 걸립니다).\n`
679
+ + ' 설치가 끝나면 다음 요청부터 자동으로 변환합니다. 이번 turn 은 원본을 그대로 읽습니다.\n'
680
+ + ` 진행 상황: ${INSTALL_HINT} 직접 실행하면 설치 로그를 수 있습니다.`,
681
+ };
682
+ }
683
+ if (result.reason === 'encrypted') {
684
+ return {
685
+ deny: false,
686
+ reason: `[doc2md] ${name} 는 암호가 걸린 문서라 변환하지 못했습니다. `
687
+ + '사용자에게 암호를 푼 사본을 요청하십시오. 이 도구는 암호를 묻거나 저장하지 않습니다.',
688
+ };
689
+ }
690
+ if (result.reason === 'python-too-old') {
691
+ if (noticeAlreadyShown()) return null;
692
+ markNoticeShown();
693
+ return {
694
+ deny: false,
695
+ reason: `[doc2md] ${name} 를 변환하지 못했습니다. 변환기(markitdown)는 Python 3.10 이상이 필요한데 PATH 에서 찾지 못했습니다.\n`
696
+ + ' macOS 기본 /usr/bin/python3 는 3.9 입니다. `brew install python` 으로 새 버전을 설치한 뒤 다시 시도하십시오.\n'
697
+ + ' 그때까지는 원본을 그대로 읽습니다. 이 안내는 한 번만 표시됩니다.',
496
698
  };
497
699
  }
498
700
  if (result.reason === 'no-text') {
@@ -613,8 +815,16 @@ function contextForPrompt(payload, opts = {}) {
613
815
  }
614
816
  } else if (result.reason === 'no-markitdown') {
615
817
  lines.push(lang === 'ko'
616
- ? ` ${name}: 변환기가 없어 변환하지 못했습니다. 설치: ${INSTALL_HINT}`
617
- : ` ${name}: no converter installed. Install it with: ${INSTALL_HINT}`);
818
+ ? ` ${name}: 변환기를 설치하는 중입니다(첫 실행에만 걸립니다). 설치가 끝나면 다음 요청부터 자동 변환됩니다.`
819
+ : ` ${name}: the converter is installing now (first run only). It will convert automatically from the next request.`);
820
+ } else if (result.reason === 'encrypted') {
821
+ lines.push(lang === 'ko'
822
+ ? ` ${name}: 암호가 걸린 문서입니다. 암호를 푼 사본을 달라고 사용자에게 요청하십시오.`
823
+ : ` ${name}: the document is password-protected. Ask the user for an unlocked copy.`);
824
+ } else if (result.reason === 'python-too-old') {
825
+ lines.push(lang === 'ko'
826
+ ? ` ${name}: 변환기가 Python 3.10 이상을 요구하는데 PATH 에 없습니다(macOS 기본은 3.9). \`brew install python\` 후 다시 시도하도록 사용자에게 안내하십시오.`
827
+ : ` ${name}: the converter needs Python 3.10+, and none is on PATH (macOS ships 3.9). Tell the user to install a newer Python, e.g. \`brew install python\`.`);
618
828
  } else if (result.reason === 'no-figparser') {
619
829
  lines.push(lang === 'ko'
620
830
  ? ` ${name}: .fig 파서가 없어 변환하지 못했습니다. 설치: claude-token-saver doc2md install-converter`
@@ -744,6 +954,10 @@ module.exports = {
744
954
  MARKITDOWN_SPEC,
745
955
  managedPython,
746
956
  installConverter,
957
+ ensureConverterInstalled,
958
+ findVenvBase,
959
+ interpreterCandidates,
960
+ interpreterParts,
747
961
  clearNotice,
748
962
  cacheDir,
749
963
  cachePathFor,
package/src/fig2md.cjs CHANGED
@@ -28,7 +28,10 @@ const fs = require('node:fs');
28
28
  const path = require('node:path');
29
29
  const { spawnSync } = require('node:child_process');
30
30
 
31
- const FIG_PARSER_SPEC = 'openfig-core@^0.4.1';
31
+ // `0.4.x` rather than `^0.4.1` on purpose: the Windows install goes through
32
+ // cmd.exe, where `^` is the escape character and would be eaten before npm
33
+ // ever saw it. The range is the same one caret means for a 0.x package.
34
+ const FIG_PARSER_SPEC = 'openfig-core@0.4.x';
32
35
 
33
36
  /** Where the on-demand parser install lives, under the tool's state dir. */
34
37
  function managedFigDir(userDataDir) {
@@ -56,10 +59,16 @@ function installFigParser(userDataDir, { onProgress = () => {} } = {}) {
56
59
  fs.writeFileSync(pkgJson, JSON.stringify({ name: 'doc2md-fig', private: true }) + '\n');
57
60
  }
58
61
  onProgress(`installing ${FIG_PARSER_SPEC}`);
59
- const r = spawnSync('npm', ['install', '--no-audit', '--no-fund', '--silent', FIG_PARSER_SPEC], {
62
+ // npm on Windows is npm.cmd, which spawnSync cannot execute directly — it
63
+ // needs the shell. Elsewhere the shell is avoided, since the package spec
64
+ // would then go through shell quoting for no benefit.
65
+ const isWindows = process.platform === 'win32';
66
+ const r = spawnSync(isWindows ? 'npm.cmd' : 'npm', ['install', '--no-audit', '--no-fund', '--silent', FIG_PARSER_SPEC], {
60
67
  cwd: dir,
61
68
  encoding: 'utf8',
62
69
  timeout: 300_000,
70
+ windowsHide: true,
71
+ shell: isWindows,
63
72
  });
64
73
  if (r.status !== 0) {
65
74
  return { ok: false, reason: 'npm-failed', detail: (r.stderr || '').slice(0, 400) };