doubao-cli 0.4.0 → 0.4.2

package/README.md

@@ -44,7 +44,11 @@ Every data-returning command supports `--json`. Select a non-default local profi
 
 Message automation requires Doubao to be launched with local Chrome DevTools Protocol enabled:
 
-Quit any running Doubao process first, then run `doubao cdp launch`. The equivalent manual command is `open -a /Applications/Doubao.app --args --remote-debugging-port=9225`.
+```bash
+doubao cdp launch
+```
+
+If Doubao is already running without CDP, the command asks for confirmation before quitting it and relaunching with the debugging port enabled. Scripts and `--json` mode never prompt; pass `doubao cdp launch --yes` to confirm the restart explicitly. The command returns only after both the CDP endpoint and authenticated chat renderer are ready. The equivalent manual sequence is to quit Doubao completely and run `open -a /Applications/Doubao.app --args --remote-debugging-port=9225`.
 
 Set `DOUBAO_CDP_ENDPOINT` if using another port. `sessions send --wait` waits for and returns the completed assistant reply.
 
@@ -93,7 +97,7 @@ No hard-coded UI coordinates, image recognition, Cookie extraction, or private c
 
 ## Limits
 
-Message read/send and attachment upload use stable DOM test ids in the authenticated Doubao renderer over localhost CDP. A Doubao update can change these selectors. The CLI verifies that uploads finish and the exact user message appears in the target conversation before reporting success. The CLI currently accepts up to 50 attachments per command and files up to 100 MiB each; the Doubao service can impose stricter type or size limits.
+Message read/send and attachment upload use stable DOM attributes in the authenticated Doubao renderer over localhost CDP. A Doubao update can change these selectors. The CLI treats image previews and file cards separately, waits for their respective upload completion signals, and verifies both the exact user message and sent attachment count before reporting success. The CLI currently accepts up to 50 attachments per command and files up to 100 MiB each; the Doubao service can impose stricter type or size limits.
 
 ## Development
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "doubao-cli",
-  "version": "0.4.0",
+  "version": "0.4.2",
   "description": "Programmatic control for local Doubao desktop sessions on macOS",
   "author": "Fullstop000 <fullstop1005@gmail.com>",
   "type": "module",

package/src/attachments.mjs

@@ -2,7 +2,9 @@ import fs from 'node:fs/promises';
 import path from 'node:path';
 
 const DROP_AREA = '[data-testid="file_drop_area"]';
+const ATTACHMENT_AREA = '[data-testid="attachment_area"]';
 const ATTACHMENT_ITEM = '[data-testid="attachment_area"] [data-testid="attachment_file_item"]';
+const IMAGE_ITEM = '[data-testid="attachment_area"] [data-testid="mdbox_image"]';
 const MAX_FILE_COUNT = 50;
 const MAX_FILE_BYTES = 100 * 1024 * 1024;
 const CHUNK_BYTES = 384 * 1024;
@@ -98,20 +100,51 @@ async function dispatchDrop(client, stateKey) {
   })()`);
 }
 
-async function waitForUploads(client, expectedCount, timeoutMs) {
+export function attachmentUploadReady(state, expected) {
+  return state.files.length === expected.files
+    && state.files.every((item) => item.available)
+    && state.images.length === expected.images
+    && state.images.every((item) => item.loaded)
+    && !state.progressing;
+}
+
+async function waitForUploads(client, files, timeoutMs) {
+  const expected = {
+    files: files.filter((file) => !file.type.startsWith('image/')).length,
+    images: files.filter((file) => file.type.startsWith('image/')).length,
+  };
   const deadline = Date.now() + timeoutMs;
   while (Date.now() < deadline) {
     const state = await client.evaluate(`(() => {
-      const items = [...document.querySelectorAll(${JSON.stringify(ATTACHMENT_ITEM)})].slice(-${expectedCount});
-      return items.map((item) => ({
-        available: item.getAttribute('data-available') === 'true',
-        name: (item.querySelector('[data-testid="message_nested_content_file_name"]')?.innerText || '').trim(),
-        status: (item.querySelector('[data-testid="message_nested_content_file_subtitle"]')?.innerText || '').trim(),
-      }));
+      const area = document.querySelector(${JSON.stringify(ATTACHMENT_AREA)});
+      const fileItems = ${expected.files}
+        ? [...document.querySelectorAll(${JSON.stringify(ATTACHMENT_ITEM)})].slice(-${expected.files})
+        : [];
+      const imageItems = ${expected.images}
+        ? [...document.querySelectorAll(${JSON.stringify(IMAGE_ITEM)})]
+          .filter((item) => !item.closest('[data-testid="attachment_file_item"]'))
+          .slice(-${expected.images})
+        : [];
+      return {
+        files: fileItems.map((item) => ({
+          available: item.getAttribute('data-available') === 'true',
+          name: (item.querySelector('[data-testid="message_nested_content_file_name"]')?.innerText || '').trim(),
+          status: (item.querySelector('[data-testid="message_nested_content_file_subtitle"]')?.innerText || '').trim(),
+        })),
+        images: imageItems.map((item) => {
+          const image = item.querySelector('img[alt="image"]');
+          return { loaded: Boolean(image?.complete && image?.naturalWidth > 0) };
+        }),
+        progressing: /(?:^|\\s)\\d{1,3}%(?:\\s|$)/u.test(area?.innerText || ''),
+        status: (area?.innerText || '').trim(),
+      };
     })()`);
-    if (state.length === expectedCount && state.every((item) => item.available)) return state;
-    const failed = state.find((item) => /(?:失败|不支持|超出|error|failed)/iu.test(item.status));
+    if (attachmentUploadReady(state, expected)) return;
+    const failed = state.files.find((item) => /(?:失败|不支持|超出|error|failed)/iu.test(item.status));
     if (failed) throw new Error(`Doubao failed to upload attachment ${failed.name || ''}: ${failed.status}`.trim());
+    if (/(?:失败|不支持|超出|error|failed)/iu.test(state.status)) {
+      throw new Error(`Doubao failed to upload an attachment: ${state.status}`);
+    }
     await delay(250);
   }
   throw new Error(`Doubao attachments did not finish uploading within ${timeoutMs} ms`);
@@ -123,13 +156,33 @@ async function clearComposerAttachments(client) {
       item.querySelector('[data-testid="message_nested_content_file_delete"]')
         ?.dispatchEvent(new MouseEvent('click', { bubbles: true, cancelable: true }));
     }
+    const area = document.querySelector(${JSON.stringify(ATTACHMENT_AREA)});
+    const imageItems = [...document.querySelectorAll(${JSON.stringify(IMAGE_ITEM)})]
+      .filter((item) => !item.closest('[data-testid="attachment_file_item"]'));
+    for (const image of imageItems) {
+      let container = image.parentElement;
+      while (container && container !== area) {
+        const deleteButton = [...container.children]
+          .find((child) => /(?:^|\\s)delete-btn-[^\\s]+/u.test(child.className || ''));
+        if (deleteButton) {
+          deleteButton.dispatchEvent(new MouseEvent('click', { bubbles: true, cancelable: true }));
+          break;
+        }
+        container = container.parentElement;
+      }
+    }
   })()`);
 }
 
 export async function uploadAttachmentsFromClient(client, filePaths, options = {}) {
   const files = await resolveAttachmentFiles(filePaths);
   if (!files.length) return [];
-  const existingAttachmentCount = await client.evaluate(`document.querySelectorAll(${JSON.stringify(ATTACHMENT_ITEM)}).length`);
+  const existingAttachmentCount = await client.evaluate(`(() => {
+    const files = document.querySelectorAll(${JSON.stringify(ATTACHMENT_ITEM)}).length;
+    const images = [...document.querySelectorAll(${JSON.stringify(IMAGE_ITEM)})]
+      .filter((item) => !item.closest('[data-testid="attachment_file_item"]')).length;
+    return files + images;
+  })()`);
   if (existingAttachmentCount) {
     throw new Error('Doubao composer already contains draft attachments; remove them before using --attach');
   }
@@ -139,19 +192,13 @@ export async function uploadAttachmentsFromClient(client, filePaths, options = {
     await stageFiles(client, files, stateKey);
     const dropped = await dispatchDrop(client, stateKey);
     if (dropped.length !== files.length) throw new Error('Doubao did not accept every attachment from the drop event');
-    let uploaded;
     try {
-      uploaded = await waitForUploads(client, files.length, timeoutMs);
+      await waitForUploads(client, files, timeoutMs);
     } catch (error) {
       await clearComposerAttachments(client).catch(() => {});
       throw error;
     }
-    return uploaded.map((item, index) => ({
-      name: item.name || files[index].name,
-      path: files[index].path,
-      size: files[index].size,
-      type: files[index].type,
-    }));
+    return files;
   } finally {
     await client.evaluate(`delete globalThis[${JSON.stringify(stateKey)}]`).catch(() => {});
   }

package/src/automation.mjs

@@ -78,8 +78,14 @@ const READ_MESSAGES_EXPRESSION = `(() => [...document.querySelectorAll('[data-te
     const attachments = [...element.querySelectorAll('[data-testid="message_nested_content_file_name"]')]
       .map((part) => (part.innerText || '').trim())
       .filter(Boolean);
-    return role && (parts.length || attachments.length)
-      ? { role, text: parts.join('\\n'), ...(attachments.length ? { attachments } : {}) }
+    const images = element.querySelectorAll('[data-plugin-identifier="block_type:10052"] img').length;
+    return role && (parts.length || attachments.length || images)
+      ? {
+        role,
+        text: parts.join('\\n'),
+        ...(attachments.length ? { attachments } : {}),
+        ...(images ? { images } : {}),
+      }
       : null;
   })
   .filter(Boolean))()`;
@@ -109,12 +115,16 @@ export function attachmentsConfirmed(before, after, attachments) {
   const beforeCounts = attachmentCounts(before);
   const afterCounts = attachmentCounts(after);
   const expectedCounts = new Map();
-  for (const attachment of attachments) {
+  for (const attachment of attachments.filter((item) => !item.type?.startsWith('image/'))) {
     expectedCounts.set(attachment.name, (expectedCounts.get(attachment.name) || 0) + 1);
   }
-  return [...expectedCounts].every(([name, expected]) => (
+  const filesConfirmed = [...expectedCounts].every(([name, expected]) => (
     (afterCounts.get(name) || 0) - (beforeCounts.get(name) || 0) >= expected
   ));
+  const expectedImages = attachments.filter((item) => item.type?.startsWith('image/')).length;
+  const beforeImages = before.reduce((total, item) => total + (item.role === 'user' ? item.images || 0 : 0), 0);
+  const afterImages = after.reduce((total, item) => total + (item.role === 'user' ? item.images || 0 : 0), 0);
+  return filesConfirmed && afterImages - beforeImages >= expectedImages;
 }
 
 export async function readConversation(id, options = {}) {

package/src/cdp.mjs

@@ -140,7 +140,7 @@ export class CdpClient {
 export async function withChatClient(callback, endpoint = cdpEndpoint()) {
   const status = await cdpStatus(endpoint);
   if (!status.available) {
-    throw new Error(`Doubao CDP is unavailable at ${endpoint}. Restart Doubao with --remote-debugging-port=9225.`);
+    throw new Error(`Doubao CDP is unavailable at ${endpoint}. Run "doubao cdp launch" to restart Doubao with CDP enabled.`);
   }
   const target = await findChatTarget(endpoint);
   const client = await new CdpClient(target.webSocketDebuggerUrl).connect();

package/src/cli.mjs

@@ -1,6 +1,7 @@
 import fs from 'node:fs';
 import path from 'node:path';
 import { spawnSync } from 'node:child_process';
+import { createInterface } from 'node:readline/promises';
 import { currentSession, getDataDir, listSessions, resolveProfile } from './storage.mjs';
 import { cdpStatus } from './cdp.mjs';
 import { createConversation, openConversation, readConversation, sendMessage } from './automation.mjs';
@@ -22,7 +23,7 @@ const HELP = `Usage:
   doubao model [--json]
   doubao model select <model> [--json]
   doubao cdp status [--json]
-  doubao cdp launch [--json]
+  doubao cdp launch [--yes] [--json]
   doubao capabilities [--json]
 
 Environment:
@@ -35,6 +36,7 @@ export function parseOptions(argv) {
   const args = [];
   let profile;
   let json = false;
+  let yes = false;
   let wait = false;
   let timeoutSeconds = 120;
   let limit = 20;
@@ -46,6 +48,8 @@ export function parseOptions(argv) {
       break;
     } else if (argv[index] === '--json') {
       json = true;
+    } else if (argv[index] === '--yes') {
+      yes = true;
     } else if (argv[index] === '--wait') {
       wait = true;
     } else if (argv[index] === '--profile') {
@@ -73,7 +77,7 @@ export function parseOptions(argv) {
       args.push(argv[index]);
     }
   }
-  return { args, profile, json, wait, timeoutMs: timeoutSeconds * 1000, limit, model, attachments };
+  return { args, profile, json, yes, wait, timeoutMs: timeoutSeconds * 1000, limit, model, attachments };
 }
 
 function output(value, json) {
@@ -89,9 +93,71 @@ function appVersion(appPath) {
   return result.status === 0 ? result.stdout.trim() : null;
 }
 
-function appRunning(appPath) {
+function appProcessPattern(appPath) {
   const executable = path.join(appPath, 'Contents', 'MacOS', 'Doubao');
-  return spawnSync('/usr/bin/pgrep', ['-f', executable]).status === 0;
+  const escaped = executable.replace(/[.*+?^${}()|[\]\\]/gu, '\\$&');
+  return `^${escaped}([[:space:]]|$)`;
+}
+
+function appRunning(appPath) {
+  return spawnSync('/usr/bin/pgrep', ['-f', appProcessPattern(appPath)]).status === 0;
+}
+
+async function waitForAppExit(appPath, timeoutMs) {
+  const deadline = Date.now() + timeoutMs;
+  while (Date.now() < deadline) {
+    if (!appRunning(appPath)) return true;
+    await new Promise((resolve) => setTimeout(resolve, 250));
+  }
+  return !appRunning(appPath);
+}
+
+async function quitAppForCdp(appPath) {
+  const result = spawnSync('/usr/bin/osascript', [
+    '-e',
+    'tell application id "com.bot.pc.doubao" to quit',
+  ], { encoding: 'utf8' });
+  if (result.status === 0 && await waitForAppExit(appPath, 5000)) return;
+
+  const terminated = spawnSync('/usr/bin/pkill', ['-TERM', '-f', appProcessPattern(appPath)], { encoding: 'utf8' });
+  if (terminated.status !== 0 && terminated.status !== 1) {
+    throw new Error(terminated.stderr.trim() || result.stderr.trim() || 'failed to stop Doubao before enabling CDP');
+  }
+  if (!await waitForAppExit(appPath, 10_000)) {
+    throw new Error('Doubao did not quit after SIGTERM. Quit it manually, then run "doubao cdp launch" again.');
+  }
+}
+
+async function confirmCdpRestart({ json, yes }) {
+  if (yes) return;
+  if (json || !process.stdin.isTTY || !process.stdout.isTTY) {
+    throw new Error('Doubao must restart to enable CDP. Re-run "doubao cdp launch --yes" to confirm.');
+  }
+
+  const prompt = createInterface({ input: process.stdin, output: process.stdout });
+  try {
+    const answer = await prompt.question('Doubao must restart to enable CDP. Continue? [y/N] ');
+    if (!/^(?:y|yes)$/iu.test(answer.trim())) {
+      throw new Error('CDP launch cancelled; Doubao was not restarted.');
+    }
+  } finally {
+    prompt.close();
+  }
+}
+
+async function waitForAutomationReady(timeoutMs = 20_000) {
+  const deadline = Date.now() + timeoutMs;
+  let lastError;
+  while (Date.now() < deadline) {
+    try {
+      const result = await listModels();
+      if (result.models.length) return;
+    } catch (error) {
+      lastError = error;
+    }
+    await new Promise((resolve) => setTimeout(resolve, 250));
+  }
+  throw new Error(`Doubao CDP is listening, but the chat renderer is not ready: ${lastError?.message || 'timed out'}`);
 }
 
 function validateId(value) {
@@ -104,7 +170,7 @@ function sessionWithTitle(profilePath, id) {
 }
 
 export async function main(argv) {
-  const { args, profile: requestedProfile, json, wait, timeoutMs, limit, model, attachments } = parseOptions(argv);
+  const { args, profile: requestedProfile, json, yes, wait, timeoutMs, limit, model, attachments } = parseOptions(argv);
   const [command, subcommand, operand] = args;
   const dataDir = getDataDir();
 
@@ -213,17 +279,22 @@ export async function main(argv) {
   if (command === 'cdp' && subcommand === 'launch') {
     const existing = await cdpStatus();
     if (existing.available) {
+      await waitForAutomationReady();
       if (json) output(existing, true);
       else console.log(`available\tyes\nendpoint\t${existing.endpoint}`);
       return;
     }
     const appPath = process.env.DOUBAO_APP || DEFAULT_APP;
-    if (appRunning(appPath)) throw new Error('Doubao is already running without CDP. Quit it completely, then run this command again.');
     const endpoint = new URL(existing.endpoint);
     if (endpoint.hostname !== '127.0.0.1' && endpoint.hostname !== 'localhost') {
       throw new Error('cdp launch only supports a localhost DOUBAO_CDP_ENDPOINT');
     }
     const port = endpoint.port || '9225';
+    const restarted = appRunning(appPath);
+    if (restarted) {
+      await confirmCdpRestart({ json, yes });
+      await quitAppForCdp(appPath);
+    }
     const result = spawnSync('/usr/bin/open', ['-a', appPath, '--args', `--remote-debugging-port=${port}`], { encoding: 'utf8' });
     if (result.status !== 0) throw new Error(result.stderr.trim() || 'failed to launch Doubao with CDP');
     let launched = existing;
@@ -232,8 +303,10 @@ export async function main(argv) {
       launched = await cdpStatus();
     }
     if (!launched.available) throw new Error(`Doubao launched, but CDP did not become available at ${existing.endpoint}`);
-    if (json) output(launched, true);
-    else console.log(`available\tyes\nendpoint\t${launched.endpoint}`);
+    await waitForAutomationReady();
+    const launchResult = { ...launched, launched: true, restarted };
+    if (json) output(launchResult, true);
+    else console.log(`available\tyes\nendpoint\t${launched.endpoint}\nrestarted\t${restarted ? 'yes' : 'no'}`);
     return;
   }