picturereader 1.0.1 → 1.0.3

package/README.md

@@ -15,7 +15,7 @@
 | 版本 | 平台 | 形态 | 源码 | 安装 |
 |---|---|---|---|---|
 | **DSH 版** | DeepSeek Harness（含 EAC 桌面端） | npm 插件（`dsh.bundle`） | 仓库根目录 | `dsh plugin --profile web add picturereader` |
-| **ZCode 版** | ZCode 桌面端 | 本地 marketplace 插件（MCP server + skill） | `zcode/` 子目录 | Settings → Plugin Management → Discover → 本地目录 |
+| **ZCode 版** | ZCode 桌面端 | 本地 marketplace 插件（MCP server + skill） | [jing-hy/picturereader-zcode](https://github.com/jing-hy/picturereader-zcode)（独立仓库） | 见其 README |
 
 两个版本共用同一套业务核心（`src/core.js`）与读图方法论 skill（`image-reading`），
 三个工具行为完全一致：`image_scan` / `image_ocr` / `image_sample`。
@@ -72,26 +72,11 @@ copy skills\image-reading.md %USERPROFILE%\.dsh\skills\   # Windows
 
 ### ZCode 版
 
-ZCode 版源码在本仓库 `zcode/` 子目录（本地开发目录为 `D:\coding\picturereader_zcode`）。
+ZCode 版是**独立仓库**：[jing-hy/picturereader-zcode](https://github.com/jing-hy/picturereader-zcode)。
 
-```text
-ZCode 桌面端 → Settings → Plugin Management → Discover → "+" → 本地目录
-→ 选择 zcode/ 目录（或本地开发目录 picturereader_zcode）
-→ 市场列表出现 picturereader → 安装
-```
-
-安装并启用后（重启 ZCode 或按提示重连）：
-
-1. **MCP server `picturereader` 自动连接**（stdio 子进程 `node mcp/server.js`），模型工具列表出现 `image_scan` / `image_ocr` / `image_sample`；
-2. **skill `image-reading` 自动出现**在技能列表（图片任务自动触发）；
-3. **（可选）PaddleOCR 增强引擎**：`node scripts/setup-ocr.mjs`。
-
-> **关于安装后的运行位置**：市场安装会把插件目录拷贝到 ZCode 插件缓存
-> （`~/.zcode/cli/plugins/cache/...`），但 `.zcode-plugin/plugin.json` 里 MCP server 的
-> `args` 是**具体绝对路径**，指向所选目录的 `mcp/server.js`——因此 MCP server 始终从
-> **所选目录**运行：改 `src/core.js` 下次调用即生效（热加载）。skill / 工具定义变更
-> 需在 Discover 里卸载后重装生效（或手动同步缓存副本）。若插件目录移动，请同步更新
-> `plugin.json` 里的 `args` 路径。
+ZCode 版通过 **MCP server**（`mcp/server.js`，stdio）把三个工具暴露给 ZCode，
+读图方法论作为 **skill**（`skills/image-reading/`）随插件分发，业务逻辑 `src/core.js`
+与本仓库完全一致。安装与使用请见其 README。
 
 ## 使用
 
@@ -123,14 +108,15 @@ luminance grid / color grid
 ## 开发
 
 ```sh
-# DSH 版（仓库根目录）
+# DSH 版（本仓库）
 npm install
 npm test            # node:test，76 个测试全绿
 node scripts/setup-ocr.mjs   # 可选：装 PaddleOCR
 node scripts/preview.mjs     # 生成 fixtures 并预览渲染
 
-# ZCode 版（zcode/ 子目录）
-cd zcode
+# ZCode 版（独立仓库 jing-hy/picturereader-zcode）
+git clone https://github.com/jing-hy/picturereader-zcode.git
+cd picturereader-zcode
 npm install
 npm test            # node:test
 node scripts/setup-ocr.mjs   # 可选
@@ -141,7 +127,7 @@ node scripts/setup-ocr.mjs   # 可选
 下次调用即生效**；工具定义（schema/描述）改动需重启桌面端。
 
 **热插拔（ZCode 版）**：MCP server 从所选目录运行，改 `src/core.js` 下次调用即生效
-（见上"安装后运行位置"）。
+（详见 picturereader-zcode 仓库 README）。
 
 ## 优势
 

package/package.json

@@ -1,6 +1,6 @@
 {
     "name":  "picturereader",
-    "version":  "1.0.1",
+    "version":  "1.0.3",
     "description":  "DSH plugin: pixel-to-text image reading for text-only models. Downscales and color-quantizes PNG/JPEG/GIF/BMP and feeds the coarse pixel grid to the model so DeepSeek can \u0027see\u0027 layout, colors and rough shapes without a vision model.",
     "type":  "module",
     "main":  "src/index.js",
@@ -11,6 +11,7 @@
                 },
     "files":  [
                   "src",
+                  "scripts",
                   "cordis.patch.yml",
                   "skills",
                   "README.md",

package/scripts/preview.mjs

@@ -0,0 +1,40 @@
+/**
+ * Generate fixture images on disk and print the image_scan render for a
+ * quick visual sanity check without booting a harness.
+ * Usage: node scripts/preview.mjs
+ * @module picturereader/scripts/preview
+ */
+
+import { mkdirSync, writeFileSync } from 'node:fs';
+import { join, dirname } from 'node:path';
+import { fileURLToPath } from 'node:url';
+import { analyzeImage, renderImageScan } from '../src/core.js';
+import { makeChartRgba, makeQuadrantRgba, pngFromRgba, jpegFromRgba, gifFromRgba, bmpFromRgba } from '../tests/fixtures.mjs';
+
+const outDir = join(dirname(fileURLToPath(import.meta.url)), '..', 'tests', 'fixtures-out');
+mkdirSync(outDir, { recursive: true });
+
+const images = [
+  { name: 'chart', width: 600, height: 400, rgba: makeChartRgba(), formats: ['png', 'jpeg', 'gif', 'bmp'] },
+  { name: 'quadrant', width: 100, height: 100, rgba: makeQuadrantRgba(), formats: ['png'] }
+];
+
+for (const image of images) {
+  for (const format of image.formats) {
+    const buffer =
+      format === 'png' ? pngFromRgba(image.width, image.height, image.rgba)
+        : format === 'jpeg' ? jpegFromRgba(image.width, image.height, image.rgba)
+          : format === 'gif' ? gifFromRgba(image.width, image.height, image.rgba)
+            : bmpFromRgba(image.width, image.height, image.rgba, 24);
+    writeFileSync(join(outDir, `${image.name}.${format}`), buffer);
+  }
+  const analysis = analyzeImage(image.rgba, image.width, image.height, { size: 32, mode: 'auto', region: undefined });
+  console.log(renderImageScan({
+    path: join(outDir, `${image.name}.png`),
+    width: image.width,
+    height: image.height,
+    region: 'full',
+    ...analysis
+  }));
+  console.log();
+}

package/scripts/setup-ocr.mjs

@@ -0,0 +1,96 @@
+/**
+ * Optional install helper for the PaddleOCR engine (image_ocr engine="paddle").
+ * PaddleOCR is RECOMMENDED (far better at glowing/curved/game text) but
+ * OPTIONAL — image_ocr degrades to the Windows engine when it is missing.
+ *
+ * What this does:
+ *  1. Ensures a Python 3.12+ interpreter exists (downloads the official
+ *     3.12.10 installer from the npmmirror mirror if missing).
+ *  2. Creates/repairs the paddle_venv.
+ *  3. Installs paddlepaddle + paddleocr from the Tsinghua PyPI mirror.
+ *  4. Warms the model cache by running one recognition on a test image.
+ *
+ * Usage: node scripts/setup-ocr.mjs
+ */
+import { existsSync, mkdirSync } from 'node:fs';
+import { spawnSync } from 'node:child_process';
+import { join, dirname } from 'node:path';
+import { fileURLToPath } from 'node:url';
+
+const PY312 = 'C:\\Users\\Administrator\\Python312\\python.exe';
+const VENV = 'C:\\Users\\Administrator\\paddle_venv\\Scripts\\python.exe';
+const INSTALLER = 'C:\\Users\\Administrator\\Downloads\\python-3.12.10-amd64.exe';
+const INSTALLER_URL = 'https://registry.npmmirror.com/-/binary/python/3.12.10/python-3.12.10-amd64.exe';
+const PYPI = 'https://pypi.tuna.tsinghua.edu.cn/simple';
+const CACHE = join(dirname(fileURLToPath(import.meta.url)), '..', '.paddlex-cache');
+
+function run(cmd, args, opts = {}) {
+  console.log(`> ${cmd} ${args.join(' ')}`);
+  const result = spawnSync(cmd, args, { stdio: 'inherit', ...opts });
+  if (result.status !== 0) {
+    console.error(`!! command failed (exit ${result.status})`);
+    process.exit(1);
+  }
+}
+
+// 1. base Python 3.12
+if (!existsSync(PY312)) {
+  console.log('[1/4] Python 3.12 missing — downloading installer (npmmirror mirror)...');
+  run('curl.exe', ['-L', '-o', INSTALLER, INSTALLER_URL]);
+  console.log('[1/4] Installing Python 3.12 to C:\\Users\\Administrator\\Python312 (user-level, silent)...');
+  spawnSync(INSTALLER, [
+    '/quiet', 'InstallAllUsers=0', 'TargetDir=C:\\Users\\Administrator\\Python312',
+    'Include_pip=1', 'PrependPath=0', 'Include_test=0', 'Include_launcher=0'
+  ], { stdio: 'inherit' });
+  if (!existsSync(PY312)) {
+    console.error('!! Python install did not produce ' + PY312);
+    process.exit(1);
+  }
+} else {
+  console.log('[1/4] Python 3.12 found at ' + PY312);
+}
+
+// 2. venv
+if (!existsSync(VENV)) {
+  console.log('[2/4] Creating paddle_venv...');
+  mkdirSync(dirname(VENV), { recursive: true });
+  run(PY312, ['-m', 'venv', 'C:\\Users\\Administrator\\paddle_venv']);
+} else {
+  console.log('[2/4] paddle_venv found');
+}
+
+// 3. paddlepaddle + paddleocr
+const probe = spawnSync(VENV, ['-c', 'import paddleocr; print(paddleocr.__version__)'], { encoding: 'utf8' });
+if (probe.status !== 0) {
+  console.log('[3/4] Installing paddlepaddle + paddleocr (Tsinghua mirror, ~1-3 min)...');
+  run(VENV, ['-m', 'pip', 'install', '-i', PYPI, '--upgrade', 'pip']);
+  run(VENV, ['-m', 'pip', 'install', '-i', PYPI, 'paddlepaddle==3.3.1', 'paddleocr']);
+} else {
+  console.log(`[3/4] paddleocr already installed (${probe.stdout.trim()})`);
+}
+
+// 4. warm the model cache with one recognition
+console.log('[4/4] Warming the model cache (first run downloads detection/recognition models)...');
+const testImage = join(dirname(fileURLToPath(import.meta.url)), '..', 'tests', 'fixtures-out', 'ocr-test.png');
+mkdirSync(join(CACHE), { recursive: true });
+if (existsSync(testImage)) {
+  const warm = spawnSync(VENV, ['-c', [
+    'from paddleocr import PaddleOCR',
+    "ocr = PaddleOCR(lang='ch', use_doc_orientation_classify=False, use_doc_unwarping=False, use_textline_orientation=False, enable_mkldnn=False)",
+    `result = ocr.predict(r'${testImage.replaceAll("'", "''")}')`,
+    'print("warm-up OCR ok, lines:", sum(len(r.get("rec_texts") or []) for r in result))'
+  ].join('; ')], {
+    env: { ...process.env, PADDLE_PDX_CACHE_HOME: CACHE, PYTHONIOENCODING: 'utf-8' },
+    encoding: 'utf8'
+  });
+  if (warm.status !== 0) {
+    console.error('!! warm-up failed — see output above; the engine may still work once models download');
+    process.exit(1);
+  }
+  console.log(warm.stdout.trim());
+} else {
+  console.log('[4/4] test image missing — skip warm-up (first image_ocr paddle call will download models)');
+}
+
+console.log('\nDone. image_ocr engine="paddle" is now available.');
+console.log('Verify: ask the model to read an image with image_ocr(engine="paddle").');