rust-rpa 0.1.7 → 0.2.0-beta.0

package/bin/rpa.js

@@ -28,7 +28,9 @@ function printTable(rows, keys) {
 }
 
 function findWindow(target) {
-  const win = Window.all().find(w =>
+  // 获取所有窗口，按面积从大到小排序，优先匹配大窗口
+  const windows = Window.all().sort((a, b) => (b.width() * b.height()) - (a.width() * a.height()));
+  const win = windows.find(w =>
     String(w.id()) === target || w.appName().includes(target) || w.title().includes(target)
   );
   if (!win) die(`未找到匹配 "${target}" 的窗口`);
@@ -247,5 +249,518 @@ clipboard.command('paste-file').description('粘贴文件（不影响剪贴板
     console.log(c.green(`已粘贴文件: ${paths.join(', ')}（剪贴板已恢复）`));
   });
 
+// ── image ──────────────────────────────────────────────────────────────────
+const image = program.command('image').description('图片处理');
+
+image.command('info').description('查看图片信息')
+  .argument('<file>', '图片文件路径')
+  .action(async (file) => {
+    const img = await ImageData.fromFile(file);
+    const meta = img.metadata();
+    console.log(`文件: ${file}`);
+    console.log(`尺寸: ${meta.width}x${meta.height}`);
+    console.log(`通道: ${meta.channels} (RGBA)`);
+    console.log(`数据大小: ${meta.dataSize} bytes`);
+  });
+
+image.command('crop').description('裁剪图片')
+  .argument('<file>', '图片文件路径')
+  .argument('<x>', '裁剪区域 X 坐标')
+  .argument('<y>', '裁剪区域 Y 坐标')
+  .argument('<width>', '裁剪区域宽度')
+  .argument('<height>', '裁剪区域高度')
+  .option('-o, --output <file>', '输出文件路径', 'cropped.png')
+  .action(async (file, x, y, width, height, { output }) => {
+    const img = await ImageData.fromFile(file);
+    const cropped = await img.crop(+x, +y, +width, +height);
+    await cropped.toFile(output);
+    console.log(c.green(`已裁剪: ${output} (${cropped.width}x${cropped.height})`));
+  });
+
+image.command('resize').description('缩放图片')
+  .argument('<file>', '图片文件路径')
+  .argument('<width>', '目标宽度')
+  .argument('<height>', '目标高度')
+  .option('-o, --output <file>', '输出文件路径', 'resized.png')
+  .action(async (file, width, height, { output }) => {
+    const img = await ImageData.fromFile(file);
+    const resized = await img.resize(+width, +height);
+    await resized.toFile(output);
+    console.log(c.green(`已缩放: ${output} (${resized.width}x${resized.height})`));
+  });
+
+image.command('grayscale').description('转换为灰度图')
+  .argument('<file>', '图片文件路径')
+  .option('-o, --output <file>', '输出文件路径', 'gray.png')
+  .action(async (file, { output }) => {
+    const img = await ImageData.fromFile(file);
+    const gray = await img.grayscale();
+    await gray.toFile(output);
+    console.log(c.green(`已转换灰度图: ${output}`));
+  });
+
+image.command('find-text').description('在图片中查找文字（OCR）')
+  .argument('<file>', '图片文件路径')
+  .argument('<text>', '要查找的文字')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (file, text, { region }) => {
+    const img = await ImageData.fromFile(file);
+    const options = {};
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    const result = await img.findText(text, options);
+    if (result) {
+      console.log(c.green(`找到文字: "${result.text}"`));
+      console.log(`位置: (${result.x}, ${result.y})`);
+      console.log(`大小: ${result.width}x${result.height}`);
+      console.log(`置信度: ${(result.confidence * 100).toFixed(2)}%`);
+    } else {
+      console.log(c.yellow('未找到文字'));
+      process.exit(1);
+    }
+  });
+
+image.command('recognize-text').description('识别图片中的所有文字（OCR）')
+  .argument('<file>', '图片文件路径')
+  .option('--json', '输出 JSON 格式')
+  .action(async (file, { json }) => {
+    const img = await ImageData.fromFile(file);
+    const results = await img.recognizeText();
+    if (json) {
+      console.log(JSON.stringify(results, null, 2));
+    } else {
+      console.log(c.green(`识别到 ${results.length} 处文字:`));
+      results.forEach((r, i) => {
+        console.log(`\n${i + 1}. "${r.text}"`);
+        console.log(`   位置: (${r.x}, ${r.y})  大小: ${r.width}x${r.height}`);
+        console.log(`   置信度: ${(r.confidence * 100).toFixed(2)}%`);
+      });
+    }
+  });
+
+image.command('find-icon').description('在图片中查找图标')
+  .argument('<file>', '图片文件路径')
+  .argument('<template>', '模板图片路径')
+  .option('--threshold <n>', '匹配阈值 (0.0-1.0)', '0.8')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (file, template, { threshold, region }) => {
+    const img = await ImageData.fromFile(file);
+    const tmpl = await ImageData.fromFile(template);
+    const options = { threshold: +threshold };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    const result = await img.findIcon(tmpl, options);
+    if (result.found) {
+      console.log(c.green(`找到图标`));
+      console.log(`位置: (${result.x}, ${result.y})`);
+      console.log(`大小: ${result.width}x${result.height}`);
+      console.log(`相似度: ${(result.score * 100).toFixed(2)}%`);
+    } else {
+      console.log(c.yellow('未找到图标'));
+      console.log(`最高相似度: ${(result.score * 100).toFixed(2)}%`);
+      process.exit(1);
+    }
+  });
+
+// ── monitor ocr ─────────────────────────────────────────────────────────────
+monitor.command('find-text').description('在显示器截图中查找文字（OCR）')
+  .argument('<text>', '要查找的文字')
+  .option('--monitor <id>', '指定显示器 ID')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (text, { monitor: monitorId, region }) => {
+    const monitors = Monitor.all();
+    let mon;
+    if (monitorId) {
+      mon = monitors.find(m => String(m.id()) === monitorId);
+      if (!mon) die(`未找到 ID 为 ${monitorId} 的显示器`);
+    } else {
+      mon = monitors.find(m => m.isPrimary()) || monitors[0];
+    }
+    const options = {};
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    const result = await mon.findText(text, options);
+    if (result) {
+      console.log(c.green(`找到文字: "${result.text}"`));
+      console.log(`屏幕位置: (${result.x}, ${result.y})`);
+      console.log(`大小: ${result.width}x${result.height}`);
+    } else {
+      console.log(c.yellow('未找到文字'));
+      process.exit(1);
+    }
+  });
+
+monitor.command('recognize-text').description('识别显示器截图中的所有文字（OCR）')
+  .option('--monitor <id>', '指定显示器 ID')
+  .option('--json', '输出 JSON 格式')
+  .action(async ({ monitor: monitorId, json }) => {
+    const monitors = Monitor.all();
+    let mon;
+    if (monitorId) {
+      mon = monitors.find(m => String(m.id()) === monitorId);
+      if (!mon) die(`未找到 ID 为 ${monitorId} 的显示器`);
+    } else {
+      mon = monitors.find(m => m.isPrimary()) || monitors[0];
+    }
+    const results = await mon.recognizeText();
+    if (json) {
+      console.log(JSON.stringify(results, null, 2));
+    } else {
+      console.log(c.green(`识别到 ${results.length} 处文字:`));
+      results.forEach((r, i) => {
+        console.log(`\n${i + 1}. "${r.text}"`);
+        console.log(`   屏幕位置: (${r.x}, ${r.y})  大小: ${r.width}x${r.height}`);
+      });
+    }
+  });
+
+monitor.command('find-icon').description('在显示器截图中查找图标')
+  .argument('<template>', '模板图片路径')
+  .option('--monitor <id>', '指定显示器 ID')
+  .option('--threshold <n>', '匹配阈值 (0.0-1.0)', '0.8')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (template, { monitor: monitorId, threshold, region }) => {
+    const monitors = Monitor.all();
+    let mon;
+    if (monitorId) {
+      mon = monitors.find(m => String(m.id()) === monitorId);
+      if (!mon) die(`未找到 ID 为 ${monitorId} 的显示器`);
+    } else {
+      mon = monitors.find(m => m.isPrimary()) || monitors[0];
+    }
+    const tmpl = await ImageData.fromFile(template);
+    const options = { threshold: +threshold };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    const result = await mon.findIcon(tmpl, options);
+    if (result.found) {
+      console.log(c.green(`找到图标`));
+      console.log(`屏幕位置: (${result.x}, ${result.y})`);
+      console.log(`大小: ${result.width}x${result.height}`);
+      console.log(`相似度: ${(result.score * 100).toFixed(2)}%`);
+    } else {
+      console.log(c.yellow('未找到图标'));
+      process.exit(1);
+    }
+  });
+
+monitor.command('wait-text').description('等待指定文字出现（OCR）')
+  .argument('<text>', '要等待的文字')
+  .option('--monitor <id>', '指定显示器 ID')
+  .option('--timeout <ms>', '超时时间（毫秒）', '3000')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (text, { monitor: monitorId, timeout, region }) => {
+    const monitors = Monitor.all();
+    let mon;
+    if (monitorId) {
+      mon = monitors.find(m => String(m.id()) === monitorId);
+      if (!mon) die(`未找到 ID 为 ${monitorId} 的显示器`);
+    } else {
+      mon = monitors.find(m => m.isPrimary()) || monitors[0];
+    }
+    const options = { timeout: +timeout };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    console.log(`等待文字 "${text}" 出现（超时: ${timeout}ms）...`);
+    const result = await mon.waitText(text, options);
+    console.log(c.green(`找到文字: "${result.text}"`));
+    console.log(`屏幕位置: (${result.x}, ${result.y})`);
+  });
+
+monitor.command('wait-icon').description('等待指定图标出现')
+  .argument('<template>', '模板图片路径')
+  .option('--monitor <id>', '指定显示器 ID')
+  .option('--timeout <ms>', '超时时间（毫秒）', '3000')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (template, { monitor: monitorId, timeout, region }) => {
+    const monitors = Monitor.all();
+    let mon;
+    if (monitorId) {
+      mon = monitors.find(m => String(m.id()) === monitorId);
+      if (!mon) die(`未找到 ID 为 ${monitorId} 的显示器`);
+    } else {
+      mon = monitors.find(m => m.isPrimary()) || monitors[0];
+    }
+    const tmpl = await ImageData.fromFile(template);
+    const options = { timeout: +timeout };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    console.log(`等待图标出现（超时: ${timeout}ms）...`);
+    const result = await mon.waitIcon(tmpl, options);
+    console.log(c.green(`找到图标`));
+    console.log(`屏幕位置: (${result.x}, ${result.y})`);
+  });
+
+monitor.command('click-text').description('点击指定文字（OCR）')
+  .argument('<text>', '要点击的文字')
+  .option('--monitor <id>', '指定显示器 ID')
+  .option('--timeout <ms>', '超时时间（毫秒）', '3000')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (text, { monitor: monitorId, timeout, region }) => {
+    const monitors = Monitor.all();
+    let mon;
+    if (monitorId) {
+      mon = monitors.find(m => String(m.id()) === monitorId);
+      if (!mon) die(`未找到 ID 为 ${monitorId} 的显示器`);
+    } else {
+      mon = monitors.find(m => m.isPrimary()) || monitors[0];
+    }
+    const options = { timeout: +timeout };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    console.log(`等待并点击文字 "${text}"（超时: ${timeout}ms）...`);
+    await mon.clickText(text, options);
+    console.log(c.green(`已点击文字`));
+  });
+
+monitor.command('click-icon').description('点击指定图标')
+  .argument('<template>', '模板图片路径')
+  .option('--monitor <id>', '指定显示器 ID')
+  .option('--timeout <ms>', '超时时间（毫秒）', '3000')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (template, { monitor: monitorId, timeout, region }) => {
+    const monitors = Monitor.all();
+    let mon;
+    if (monitorId) {
+      mon = monitors.find(m => String(m.id()) === monitorId);
+      if (!mon) die(`未找到 ID 为 ${monitorId} 的显示器`);
+    } else {
+      mon = monitors.find(m => m.isPrimary()) || monitors[0];
+    }
+    const tmpl = await ImageData.fromFile(template);
+    const options = { timeout: +timeout };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    console.log(`等待并点击图标（超时: ${timeout}ms）...`);
+    await mon.clickIcon(tmpl, options);
+    console.log(c.green(`已点击图标`));
+  });
+
+// ── window ocr ──────────────────────────────────────────────────────────────
+window.command('capture').description('截取窗口截图')
+  .argument('<target>', '按 appName/title/id 匹配')
+  .option('-o, --output <file>', '输出文件路径', 'window-capture.png')
+  .action(async (target, { output }) => {
+    const win = findWindow(target);
+    console.log(`截取窗口: ${win.appName()} - ${win.title()}`);
+    const image = await win.captureImage();
+    await image.toFile(output);
+    console.log(c.green(`已保存: ${output} (${image.width}x${image.height})`));
+  });
+
+window.command('find-text').description('在窗口截图中查找文字（OCR）')
+  .argument('<target>', '按 appName/title/id 匹配')
+  .argument('<text>', '要查找的文字')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (target, text, { region }) => {
+    const win = findWindow(target);
+    const options = {};
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    const result = await win.findText(text, options);
+    if (result) {
+      console.log(c.green(`找到文字: "${result.text}"`));
+      console.log(`窗口内位置: (${result.x}, ${result.y})`);
+      console.log(`大小: ${result.width}x${result.height}`);
+    } else {
+      console.log(c.yellow('未找到文字'));
+      process.exit(1);
+    }
+  });
+
+window.command('recognize-text').description('识别窗口截图中的所有文字（OCR）')
+  .argument('<target>', '按 appName/title/id 匹配')
+  .option('--json', '输出 JSON 格式')
+  .action(async (target, { json }) => {
+    const win = findWindow(target);
+    const results = await win.recognizeText();
+    if (json) {
+      console.log(JSON.stringify(results, null, 2));
+    } else {
+      console.log(c.green(`识别到 ${results.length} 处文字:`));
+      results.forEach((r, i) => {
+        console.log(`\n${i + 1}. "${r.text}"`);
+        console.log(`   窗口内位置: (${r.x}, ${r.y})  大小: ${r.width}x${r.height}`);
+      });
+    }
+  });
+
+window.command('find-icon').description('在窗口截图中查找图标')
+  .argument('<target>', '按 appName/title/id 匹配')
+  .argument('<template>', '模板图片路径')
+  .option('--threshold <n>', '匹配阈值 (0.0-1.0)', '0.8')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (target, template, { threshold, region }) => {
+    const win = findWindow(target);
+    const tmpl = await ImageData.fromFile(template);
+    const options = { threshold: +threshold };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    const result = await win.findIcon(tmpl, options);
+    if (result.found) {
+      console.log(c.green(`找到图标`));
+      console.log(`窗口内位置: (${result.x}, ${result.y})`);
+      console.log(`大小: ${result.width}x${result.height}`);
+      console.log(`相似度: ${(result.score * 100).toFixed(2)}%`);
+    } else {
+      console.log(c.yellow('未找到图标'));
+      process.exit(1);
+    }
+  });
+
+window.command('wait-text').description('等待窗口内指定文字出现（OCR）')
+  .argument('<target>', '按 appName/title/id 匹配')
+  .argument('<text>', '要等待的文字')
+  .option('--timeout <ms>', '超时时间（毫秒）', '3000')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (target, text, { timeout, region }) => {
+    const win = findWindow(target);
+    const options = { timeout: +timeout };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    console.log(`等待文字 "${text}" 出现（超时: ${timeout}ms）...`);
+    const result = await win.waitText(text, options);
+    console.log(c.green(`找到文字: "${result.text}"`));
+    console.log(`屏幕位置: (${result.x}, ${result.y})`);
+  });
+
+window.command('wait-icon').description('等待窗口内指定图标出现')
+  .argument('<target>', '按 appName/title/id 匹配')
+  .argument('<template>', '模板图片路径')
+  .option('--timeout <ms>', '超时时间（毫秒）', '3000')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (target, template, { timeout, region }) => {
+    const win = findWindow(target);
+    const tmpl = await ImageData.fromFile(template);
+    const options = { timeout: +timeout };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    console.log(`等待图标出现（超时: ${timeout}ms）...`);
+    const result = await win.waitIcon(tmpl, options);
+    console.log(c.green(`找到图标`));
+    console.log(`屏幕位置: (${result.x}, ${result.y})`);
+  });
+
+window.command('click-text').description('点击窗口内指定文字（OCR）')
+  .argument('<target>', '按 appName/title/id 匹配')
+  .argument('<text>', '要点击的文字')
+  .option('--timeout <ms>', '超时时间（毫秒）', '3000')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (target, text, { timeout, region }) => {
+    const win = findWindow(target);
+    const options = { timeout: +timeout };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    console.log(`等待并点击文字 "${text}"（超时: ${timeout}ms）...`);
+    await win.clickText(text, options);
+    console.log(c.green(`已点击文字`));
+  });
+
+window.command('click-icon').description('点击窗口内指定图标')
+  .argument('<target>', '按 appName/title/id 匹配')
+  .argument('<template>', '模板图片路径')
+  .option('--timeout <ms>', '超时时间（毫秒）', '3000')
+  .option('--region <x,y,w,h>', '指定查找区域，格式: x,y,w,h')
+  .action(async (target, template, { timeout, region }) => {
+    const win = findWindow(target);
+    const tmpl = await ImageData.fromFile(template);
+    const options = { timeout: +timeout };
+    if (region) {
+      const [x, y, w, h] = region.split(',').map(Number);
+      options.regions = [{ x, y, width: w, height: h }];
+    }
+    console.log(`等待并点击图标（超时: ${timeout}ms）...`);
+    await win.clickIcon(tmpl, options);
+    console.log(c.green(`已点击图标`));
+  });
+
+// ── shortcuts ───────────────────────────────────────────────────────────────
+// 常用命令的顶级快捷方式（无需指定命名空间）
+
+program.command('paste-text').description('粘贴文本（快捷方式: clipboard paste-text）')
+  .argument('<text...>', '要粘贴的文本')
+  .action(async (parts) => {
+    const text = parts.join(' ');
+    await Clipboard.pasteText(text);
+    console.log(c.green(`已粘贴文本: "${text}"（剪贴板已恢复）`));
+  });
+
+program.command('paste').description('执行粘贴 (Cmd/Ctrl+V)（快捷方式: clipboard paste）')
+  .action(async () => {
+    await Clipboard.paste();
+    console.log(c.green('已执行粘贴'));
+  });
+
+program.command('type').description('输入文本（快捷方式: keyboard type）')
+  .argument('<text...>', '要输入的文本')
+  .action(async (parts) => {
+    const text = parts.join(' ');
+    await Keyboard.typeText(text);
+    console.log(c.green(`已输入: "${text}"`));
+  });
+
+program.command('press').description('按键或组合键（快捷方式: keyboard press）')
+  .argument('<keys>', '按键（组合键用 + 连接，如 ctrl+c）')
+  .action(async (raw) => {
+    const keys = raw.split('+').map(k => k.trim().toLowerCase());
+    if (keys.length === 1) await Keyboard.click(keys[0]);
+    else await Keyboard.sequence(keys);
+    console.log(c.green(`已按下: ${raw}`));
+  });
+
+program.command('click').description('点击鼠标（快捷方式: mouse click）')
+  .argument('[button]', '按钮 (left/right/middle)', 'left')
+  .option('--double', '双击')
+  .action(async (button, { double }) => {
+    if (double) {
+      await Mouse.doubleClick(button);
+      console.log(c.green(`双击: ${button}`));
+    } else {
+      await Mouse.click(button);
+      console.log(c.green(`点击: ${button}`));
+    }
+  });
+
+program.command('move').description('移动鼠标（快捷方式: mouse move）')
+  .argument('<x>', 'X 坐标').argument('<y>', 'Y 坐标')
+  .action(async (x, y) => {
+    await Mouse.moveTo(+x, +y);
+    console.log(c.green(`鼠标已移动到 (${x}, ${y})`));
+  });
+
+program.command('pause').description('暂停/等待指定时间（毫秒）')
+  .argument('<ms>', '等待时间（毫秒）')
+  .action(async (ms) => {
+    const delay = parseInt(ms, 10);
+    await new Promise(resolve => setTimeout(resolve, delay));
+    console.log(c.green(`已等待 ${delay}ms`));
+  });
+
 // ── run ─────────────────────────────────────────────────────────────────────
 program.parseAsync().catch(err => die(err.message));