mu-agent-bridge 1.0.34 → 1.0.35

package/dist/src-ts/plugins/chatgpt/cdp-client.js

@@ -1,9 +1,45 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.ChatgptCdpClient = exports.userBubbleSelector = exports.turnSelector = exports.editorSelector = void 0;
+exports.chatgptAttachmentPath = chatgptAttachmentPath;
 const node_child_process_1 = require("node:child_process");
+const node_crypto_1 = require("node:crypto");
+const attachments_1 = require("../../attachments");
 const session_store_1 = require("./session-store");
 const cdp_draft_1 = require("../cdp-draft");
+function chatgptAttachmentPath(href, outputDirectory) {
+    let path;
+    try {
+        if (decodeURIComponent(href).split(/[\\/]/).some(part => part === "." || part === ".."))
+            return;
+        if (href.startsWith("file://")) {
+            const url = new URL(href);
+            if (url.hostname || url.search || url.hash || url.username || url.password)
+                return;
+            path = decodeURIComponent(url.pathname);
+        }
+        else if (href.startsWith("sandbox:/"))
+            path = decodeURIComponent(href.slice("sandbox:".length));
+        else if (href.startsWith("/") || /^[a-z]:\//i.test(href))
+            path = decodeURIComponent(href);
+        else
+            return;
+    }
+    catch {
+        return;
+    }
+    if (/[%?#\u0000-\u001f\u007f]/.test(path) || path.startsWith("//") || path.includes("\\"))
+        return;
+    if (path.split("/").some(part => part === "." || part === ".."))
+        return;
+    const normalized = /^\/[a-z]:\//i.test(path) ? path.slice(1) : path;
+    const unixRoot = `/tmp/${outputDirectory}/`;
+    const windowsRoot = normalized.match(/^[a-z]:\/Users\/[^/]+\/AppData\/Local\/Temp\//i)?.[0];
+    const prefix = normalized.startsWith(unixRoot) ? unixRoot : windowsRoot && normalized.startsWith(`${windowsRoot}${outputDirectory}/`) ? `${windowsRoot}${outputDirectory}/` : undefined;
+    if (!prefix || normalized.slice(prefix.length).split("/").some(part => !part || part.startsWith(".")))
+        return;
+    return normalized;
+}
 exports.editorSelector = '[data-codex-composer="true"][contenteditable="true"]';
 exports.turnSelector = '[data-chatgpt-conversation-turn="true"]';
 exports.userBubbleSelector = "[data-user-message-bubble]";
@@ -16,6 +52,7 @@ const sleep = (ms) => new Promise((resolve) => setTimeout(resolve, ms));
 const sidebarSelector = "[data-app-action-sidebar-thread-id]";
 const newSessionSelector = 'button,[role="button"]';
 class ChatgptCdpClient {
+    attachmentInput = true;
     socket;
     tunnel;
     sequence = 0;
@@ -190,15 +227,55 @@ class ChatgptCdpClient {
         this.tunnel = undefined;
         this.rejectPending(new Error("ChatGPT CDP 已关闭"));
     }
-    async direct(text, timeoutMs = 240_000) {
+    async direct(text, timeoutMs = 240_000, attachments) {
+        (0, attachments_1.validateAttachments)(attachments);
+        const slot = `__agentBridgeChatgpt_${(0, node_crypto_1.randomUUID)().replaceAll("-", "")}`;
+        const outputDirectory = `agent-bridge-chatgpt-${(0, node_crypto_1.randomUUID)()}`;
+        const startedAt = Date.now();
+        const instruction = `附件回传约定：如果本轮需要返回文件、图片或视频，请把要返回的文件副本写入本轮新建的临时目录（Linux/macOS：/tmp/${outputDirectory}；Windows：C:/Users/<当前用户>/AppData/Local/Temp/${outputDirectory}），并在最终回复给出文件链接。仅需文字时无需创建目录。`;
+        let failure;
+        try {
+            return await this.directTurn(`${text} ${instruction}`, timeoutMs, attachments, slot, outputDirectory, startedAt);
+        }
+        catch (error) {
+            failure = error;
+            throw error;
+        }
+        finally {
+            if (attachments?.length) {
+                try {
+                    await this.evaluate(`(async () => {
+          const state = globalThis[${JSON.stringify(slot)}];
+          if (!state) return;
+          state.observer?.disconnect();
+          for (const button of state.buttons ?? []) if (button.isConnected) button.click();
+          try { await Promise.all(state.paths.map(path => state.services.workspaceFiles.releaseTemporaryFile({ path }))); }
+          finally { delete globalThis[${JSON.stringify(slot)}]; }
+        })()`);
+                }
+                catch (cleanupError) {
+                    if (failure === undefined)
+                        throw cleanupError;
+                    throw new AggregateError([failure, cleanupError], `${failure instanceof Error ? failure.message : String(failure)}；临时附件清理失败`, { cause: failure });
+                }
+            }
+        }
+    }
+    async directTurn(text, timeoutMs, attachments, slot, outputDirectory, startedAt) {
         await this.waitFor(`Boolean(document.querySelector(${JSON.stringify(exports.editorSelector)}))`, "ChatGPT 桌面端输入框未就绪");
         await this.recoverDraft();
         const beforeTurnsJson = await this.evaluate(`JSON.stringify([...document.querySelectorAll(${JSON.stringify(exports.turnSelector)})].map(turn => ({ id: turn.getAttribute(${JSON.stringify(turnIdAttribute)}), users: turn.querySelectorAll(${JSON.stringify(exports.userBubbleSelector)}).length })))`);
         const beforeTurns = JSON.parse(String(beforeTurnsJson || "[]"));
         const beforeAssistantCount = Number(await this.evaluate(`document.querySelectorAll(${JSON.stringify(assistantMessageSelector)}).length`));
+        const baseline = await this.evaluate(`JSON.stringify([...document.querySelectorAll(${JSON.stringify(assistantMessageSelector)})].map(node => {
+      const message = node.closest('[data-response-annotation-target][data-response-annotation-conversation]');
+      return message ? [message.getAttribute('data-response-annotation-conversation'), message.getAttribute('data-response-annotation-target')] : null;
+    }).filter(Boolean))`);
+        const beforeMessages = JSON.parse(String(baseline || "[]"));
+        const origin = await this.activeSession();
         const ready = await this.evaluate(`(() => {
       const editor = document.querySelector(${JSON.stringify(exports.editorSelector)});
-      if (!editor || editor.innerText.trim()) return false;
+      if (!editor || editor.innerText.trim() || document.querySelector('button[aria-label^="Remove "]')) return false;
       const findButton = () => {
         for (let parent = editor; parent && parent !== document.body; parent = parent.parentElement) {
           const buttons = [...parent.querySelectorAll(${JSON.stringify(sendButtonSelector)})];
@@ -213,6 +290,8 @@ class ChatgptCdpClient {
     })()`);
         if (!ready)
             throw new Error("ChatGPT 输入框含有未发送内容，或桌面端发送按钮未就绪");
+        if (attachments?.length)
+            await this.injectAttachments(attachments, slot);
         await this.cdp("Input.insertText", { text });
         const inputMatches = await this.evaluate(`document.querySelector(${JSON.stringify(exports.editorSelector)})?.innerText.trim() === ${JSON.stringify(text.trim())}`);
         if (!inputMatches)
@@ -239,22 +318,36 @@ class ChatgptCdpClient {
         while (Date.now() < deadline) {
             await sleep(800);
             const raw = await this.evaluate(`(() => {
+        const previous = new Set(${JSON.stringify(Array.isArray(beforeMessages) ? beforeMessages : [])}.map(item => JSON.stringify(item)));
+        let hasArtifacts = false;
+        const artifacts = [...document.querySelectorAll(${JSON.stringify(assistantMessageSelector)})].flatMap(node => {
+          const message = node.closest('[data-response-annotation-target][data-response-annotation-conversation]');
+          if (!message) return [];
+          const conversationId = message.getAttribute('data-response-annotation-conversation');
+          const messageId = message.getAttribute('data-response-annotation-target');
+          if (!conversationId || !messageId || previous.has(JSON.stringify([conversationId, messageId]))) return [];
+          hasArtifacts ||= Boolean(node.querySelector('a[href], img[src], video[src], source[src], [data-file-reference][data-prompt-link-href]'));
+          return [{ conversationId, messageId }];
+        });
         const before = ${JSON.stringify(beforeTurns)};
         const turns = [...document.querySelectorAll(${JSON.stringify(exports.turnSelector)})];
+        const userText = node => (node?.querySelector('[data-markdown-text-tone="user-message"]') ?? node)?.innerText.trim().replace(/\\s+/g, ' ');
+        const expectedUserText = ${JSON.stringify(text.trim().replace(/\s+/g, " "))};
         const turn = turns.findLast((item, index) => {
           const id = item.getAttribute(${JSON.stringify(turnIdAttribute)});
           const oldIndex = before.findIndex(entry => entry.id === id);
           const old = oldIndex < 0 ? undefined : before[oldIndex];
           const users = item.querySelectorAll(${JSON.stringify(exports.userBubbleSelector)});
           const fresh = oldIndex < 0 || index >= before.length || users.length > (old?.users ?? 0);
-          return fresh && users.item(users.length - 1)?.innerText.trim() === ${JSON.stringify(text.trim())};
+          return fresh && userText(users.item(users.length - 1)) === expectedUserText;
         });
         if (!turn) {
           const assistants = [...document.querySelectorAll(${JSON.stringify(assistantMessageSelector)})];
           const fallback = assistants.length > ${beforeAssistantCount}
             ? assistants.at(-1)?.innerText.trim() ?? ""
             : "";
-          return JSON.stringify({ reply: fallback, active: Boolean(document.querySelector(${JSON.stringify(stopButtonSelector)})) });
+          const submitted = [...document.querySelectorAll(${JSON.stringify(exports.userBubbleSelector)})].some(node => userText(node) === expectedUserText);
+          return JSON.stringify({ reply: fallback, artifacts, hasArtifacts, submitted, active: Boolean(document.querySelector(${JSON.stringify(stopButtonSelector)})) });
         }
         const user = turn.querySelectorAll(${JSON.stringify(exports.userBubbleSelector)}).item(turn.querySelectorAll(${JSON.stringify(exports.userBubbleSelector)}).length - 1);
         const unit = user?.closest(${JSON.stringify(contentUnitSelector)});
@@ -264,14 +357,20 @@ class ChatgptCdpClient {
         const legacyReplies = [...turn.querySelectorAll('[data-message-author-role="assistant"]')].map(item => item.innerText.trim()).filter(Boolean);
         const reply = replies.at(-1) ?? legacyReplies.at(-1) ?? "";
         const active = Boolean(document.querySelector(${JSON.stringify(stopButtonSelector)}));
-        return JSON.stringify({ reply, active });
+        return JSON.stringify({ reply, active, artifacts, hasArtifacts, submitted: true });
       })()`);
             const state = JSON.parse(String(raw));
             const reply = state.reply.trim();
-            stable = reply && reply === previous ? stable + 1 : 0;
-            previous = reply;
-            if (reply && stable >= 2 && !state.active)
-                return reply;
+            const signature = JSON.stringify([reply, state.artifacts, state.hasArtifacts]);
+            stable = (reply || state.hasArtifacts) && signature === previous ? stable + 1 : 0;
+            previous = signature;
+            if ((reply || state.hasArtifacts) && stable >= 2 && !state.active) {
+                const output = state.hasArtifacts && state.artifacts?.length ? await this.readArtifacts(state.artifacts, origin, outputDirectory, startedAt, state.submitted) : [];
+                if (output.length)
+                    return { text: reply, attachments: output };
+                if (reply)
+                    return reply;
+            }
         }
         throw new Error("ChatGPT UI 回复超时");
     }
@@ -315,6 +414,168 @@ class ChatgptCdpClient {
     recoverDraft() {
         return (0, cdp_draft_1.recoverCdpDraft)(this.evaluate.bind(this), exports.editorSelector, stopButtonSelector, this.dataDir);
     }
+    async injectAttachments(items, slot) {
+        const files = items.map(item => ({ filename: (0, attachments_1.attachmentFilename)(item.filename), contentType: item.contentType, data: Buffer.from(item.data).toString("base64") }));
+        await this.evaluate(String.raw `(async () => {
+      const editor = document.querySelector(${JSON.stringify(exports.editorSelector)});
+      if (!editor || document.querySelector('button[aria-label^="Remove "]')) throw new Error('ChatGPT 输入框已有附件或未就绪，未发送消息');
+      const entry = [...document.scripts].find(script => /\/assets\/index-[^/]+\.js$/.test(script.src));
+      if (!entry) throw new Error('ChatGPT 桌面原生接口入口未找到');
+      const source = await (await fetch(entry.src)).text();
+      const rpc = source.match(/["'](\.\/rpc-[\w-]+\.js)["']/);
+      if (!rpc) throw new Error('ChatGPT 桌面原生附件接口未找到');
+      const { appServices: services } = await import(new URL(rpc[1], entry.src).href);
+      const state = { services, paths: [], buttons: new Set(), observer: null }; globalThis[${JSON.stringify(slot)}] = state;
+      const transfer = new DataTransfer();
+      for (const item of ${JSON.stringify(files)}) {
+        const binary = atob(item.data); const bytes = new Uint8Array(binary.length);
+        for (let index = 0; index < binary.length; index++) bytes[index] = binary.charCodeAt(index);
+        const result = await services.workspaceFiles.createTemporaryFile({ bytes, fileName: item.filename });
+        if (!result?.path) throw new Error('ChatGPT 临时附件创建失败');
+        state.paths.push(result.path);
+        const file = new File([bytes], item.filename, { type: item.contentType });
+        Object.defineProperty(file, 'path', { value: result.path }); transfer.items.add(file);
+      }
+      const names = ${JSON.stringify(files.map(file => `Remove ${file.filename}`))};
+      const capture = () => {
+        for (const button of document.querySelectorAll('button[aria-label^="Remove "]')) {
+          if (names.includes(button.getAttribute('aria-label'))) state.buttons.add(button);
+        }
+      };
+      state.observer = new MutationObserver(capture);
+      state.observer.observe(document.body, { childList: true, subtree: true });
+      editor.dispatchEvent(new ClipboardEvent('paste', { bubbles: true, cancelable: true, clipboardData: transfer }));
+      capture();
+    })()`);
+        await this.waitFor(`(() => {
+      const names = ${JSON.stringify(files.map(file => `Remove ${file.filename}`))};
+      const current = [...document.querySelectorAll('button[aria-label^="Remove "]')].map(node => node.getAttribute('aria-label'));
+      const ready = current.length === names.length && names.every(name => current.filter(value => value === name).length === names.filter(value => value === name).length);
+      if (ready) globalThis[${JSON.stringify(slot)}]?.observer?.disconnect();
+      return ready;
+    })()`, "ChatGPT 附件上传未确认");
+    }
+    async readArtifacts(messages, origin, outputDirectory, startedAt, submitted = false) {
+        const active = await this.activeSession();
+        if (!origin || !active || active.host !== origin.host)
+            throw new Error("ChatGPT 附件所属会话无法确认");
+        if (active.id !== origin.id)
+            throw new Error("ChatGPT 会话已变化，未读取输出附件");
+        let conversationId = active.id.replace(/^local:/, "");
+        if (conversationId.startsWith("client-new-thread:")) {
+            const ids = [...new Set(messages.map(message => message.conversationId))];
+            if (active.id !== origin.id || !submitted || ids.length !== 1 || !/^[a-f0-9]{8}(?:-[a-f0-9]{4}){3}-[a-f0-9]{12}$/i.test(ids[0]))
+                throw new Error("ChatGPT 新会话身份尚未确认，未读取输出附件");
+            conversationId = ids[0];
+        }
+        const scoped = messages.filter(message => typeof message.conversationId === "string" && typeof message.messageId === "string"
+            && message.conversationId === conversationId);
+        const sessionGuard = `const selected = document.querySelector('[data-app-action-sidebar-thread-selected="true"], [data-app-action-sidebar-thread-active="true"]');
+      if (selected?.getAttribute('data-app-action-sidebar-thread-id') !== ${JSON.stringify(active.id)}
+        || selected?.getAttribute('data-app-action-sidebar-thread-host-id') !== ${JSON.stringify(active.host)}) throw new Error('ChatGPT 会话已变化，未读取输出附件');`;
+        const references = await this.evaluate(`(() => {
+      ${sessionGuard}
+      const messages = ${JSON.stringify(scoped)};
+      const references = [];
+      for (const node of document.querySelectorAll(${JSON.stringify(assistantMessageSelector)})) {
+        const message = node.closest('[data-response-annotation-target][data-response-annotation-conversation]');
+        if (!message || !messages.some(item => item.conversationId === message.getAttribute('data-response-annotation-conversation') && item.messageId === message.getAttribute('data-response-annotation-target'))) continue;
+        for (const link of node.querySelectorAll('a[href], img[src], video[src], source[src], [data-file-reference][data-prompt-link-href]')) {
+          const href = link.getAttribute('href') ?? link.getAttribute('src') ?? link.getAttribute('data-prompt-link-href');
+          if (href) references.push(href);
+        }
+      }
+      return references;
+    })()`);
+        const paths = [...new Set(references.map(href => chatgptAttachmentPath(href, outputDirectory)).filter((path) => Boolean(path)))];
+        if (paths.length > attachments_1.MAX_ATTACHMENTS)
+            throw new Error("单条消息最多包含 8 个附件");
+        const output = [];
+        let total = 0;
+        for (const path of paths) {
+            const fileSlot = `__agentBridgeChatgptOutput_${(0, node_crypto_1.randomUUID)().replaceAll("-", "")}`;
+            let file;
+            let readFailure;
+            try {
+                const metadata = await this.evaluate(`(async () => {
+        ${sessionGuard}
+        globalThis[${JSON.stringify(fileSlot)}] = '';
+        const call = (name, params) => new Promise((resolve, reject) => {
+          const requestId = crypto.randomUUID();
+          const finish = () => { clearTimeout(timer); window.removeEventListener('message', receive); };
+          const receive = event => {
+            const response = event.data;
+            if ((event.source !== window && !(event.source === null && event.origin === '')) || response?.type !== 'fetch-response' || response.requestId !== requestId) return;
+            finish();
+            if (response.responseType !== 'success' || response.status !== 200) { reject(new Error('ChatGPT 输出附件读取失败')); return; }
+            try { resolve(JSON.parse(response.bodyJsonString)); } catch { reject(new Error('ChatGPT 输出附件响应无效')); }
+          };
+          const timer = setTimeout(() => { finish(); reject(new Error('ChatGPT 输出附件读取超时')); }, ${this.timeoutMs});
+          window.addEventListener('message', receive);
+          try { window.electronBridge.sendMessageFromView({ type: 'fetch', requestId, method: 'POST', url: 'vscode://codex/' + name, body: JSON.stringify(params) }); }
+          catch (error) { finish(); reject(error); }
+        });
+        const params = { hostId: ${JSON.stringify(active.host)}, path: ${JSON.stringify(path)} };
+        const metadata = await call('read-file-metadata', params);
+        if (!metadata.isFile || !Number.isSafeInteger(metadata.sizeBytes) || metadata.sizeBytes <= 0) throw new Error('ChatGPT 输出附件元数据无效');
+        if (!Number.isFinite(metadata.createdAtMs) || metadata.createdAtMs < ${startedAt - 5000}) throw new Error('ChatGPT 输出附件不是本轮创建的文件');
+        if (metadata.sizeBytes > ${attachments_1.MAX_ATTACHMENT_BYTES}) throw new Error('单个附件不能超过 32 MiB');
+        if (metadata.sizeBytes > ${attachments_1.MAX_TOTAL_ATTACHMENT_BYTES - total}) throw new Error('单条消息附件总大小不能超过 64 MiB');
+        { ${sessionGuard} }
+        const entry = [...document.scripts].find(script => /\\/assets\\/index-[^/]+\\.js$/.test(script.src));
+        if (!entry) throw new Error('ChatGPT 桌面原生接口入口未找到');
+        const source = await (await fetch(entry.src)).text();
+        const rpc = source.match(/["'](\\.\\/rpc-[\\w-]+\\.js)["']/);
+        if (!rpc) throw new Error('ChatGPT 桌面原生附件接口未找到');
+        const { appServices: services } = await import(new URL(rpc[1], entry.src).href);
+        { ${sessionGuard} }
+        const file = await services.workspaceFiles.read({ ...params, representation: 'blob' });
+        { ${sessionGuard} }
+        if (globalThis[${JSON.stringify(fileSlot)}] !== '') throw new Error('ChatGPT 输出附件读取已取消');
+        if (typeof file.blob !== 'string' || file.blob.length !== Math.ceil(metadata.sizeBytes / 3) * 4) throw new Error('ChatGPT 输出附件内容无效');
+        globalThis[${JSON.stringify(fileSlot)}] = file.blob;
+        return { length: file.blob.length, size: metadata.sizeBytes };
+      })()`);
+                if (!Number.isSafeInteger(metadata.length) || metadata.length <= 0 || metadata.length > Math.ceil(attachments_1.MAX_ATTACHMENT_BYTES / 3) * 4)
+                    throw new Error("ChatGPT 输出附件内容无效");
+                const chunks = [];
+                for (let offset = 0; offset < metadata.length; offset += 256 * 1024) {
+                    const chunk = await this.evaluate(`(() => { ${sessionGuard} return globalThis[${JSON.stringify(fileSlot)}]?.slice(${offset}, ${offset + 256 * 1024}); })()`);
+                    if (typeof chunk !== "string" || chunk.length !== Math.min(256 * 1024, metadata.length - offset))
+                        throw new Error("ChatGPT 输出附件分块读取失败");
+                    chunks.push(chunk);
+                }
+                file = { ...metadata, contentsBase64: chunks.join("") };
+            }
+            catch (error) {
+                readFailure = error;
+                throw error;
+            }
+            finally {
+                try {
+                    await this.evaluate(`delete globalThis[${JSON.stringify(fileSlot)}]`);
+                }
+                catch (cleanupError) {
+                    if (readFailure === undefined)
+                        throw cleanupError;
+                    throw new AggregateError([readFailure, cleanupError], `${readFailure instanceof Error ? readFailure.message : String(readFailure)}；输出附件缓存清理失败`, { cause: readFailure });
+                }
+            }
+            if (typeof file.contentsBase64 !== "string" || file.contentsBase64.length > Math.ceil(attachments_1.MAX_ATTACHMENT_BYTES / 3) * 4)
+                throw new Error("ChatGPT 输出附件内容无效");
+            const data = Buffer.from(file.contentsBase64, "base64");
+            if (data.toString("base64") !== file.contentsBase64 || data.byteLength !== file.size)
+                throw new Error("ChatGPT 输出附件大小或编码不一致");
+            const filename = (0, attachments_1.attachmentFilename)(path);
+            const imageType = (0, attachments_1.detectImageContentType)(data);
+            const types = { txt: "text/plain", md: "text/markdown", pdf: "application/pdf", json: "application/json", zip: "application/zip", mp4: "video/mp4", webm: "video/webm", mov: "video/quicktime", mp3: "audio/mpeg", wav: "audio/wav" };
+            const contentType = imageType ?? file.mimeType ?? types[filename.split(".").at(-1)?.toLowerCase() ?? ""] ?? "application/octet-stream";
+            output.push({ kind: imageType ? "image" : "file", data, filename, contentType });
+            total += data.byteLength;
+            (0, attachments_1.validateAttachments)(output);
+        }
+        return output;
+    }
     async waitFor(expression, message) {
         const deadline = Date.now() + this.timeoutMs;
         while (Date.now() < deadline) {

package/dist/src-ts/plugins/chatgpt/index.js

@@ -5,7 +5,7 @@ const cdp_client_1 = require("./cdp-client");
 const session_store_1 = require("./session-store");
 exports.chatgptManifest = {
     id: "chatgpt",
-    version: "1.0.6",
+    version: "1.0.8",
     displayName: "ChatGPT（Codex）",
     type: "agent",
     entry: "dist/index.js",
@@ -23,6 +23,7 @@ class ChatgptPlugin {
     options;
     manifest = exports.chatgptManifest;
     sessionOperations = [];
+    get attachmentInput() { return this.client?.attachmentInput === true; }
     client;
     failure;
     remote = false;
@@ -94,7 +95,8 @@ class ChatgptPlugin {
             throw new Error("ChatGPT 插件未启动");
         const run = this.queue.then(() => this.run(request));
         this.queue = run.then(() => undefined, () => undefined);
-        return { text: await run, sessionId: request.sessionId };
+        const result = await run;
+        return { ...(typeof result === "string" ? { text: result } : result), sessionId: request.sessionId };
     }
     async controlSession(request) {
         if (request.operation !== "new")
@@ -204,7 +206,7 @@ class ChatgptPlugin {
             await this.client.connect();
         selected ??= await this.client.activeSession?.();
         const message = explicit?.[2]?.trim() || request.task;
-        const reply = await this.client.direct(message, request.timeoutMs);
+        const reply = await this.client.direct(message, request.timeoutMs, request.attachments);
         const baseline = request.sessionId ? store.pendingBindings[request.sessionId] : undefined;
         if (baseline) {
             try {
@@ -226,7 +228,7 @@ class ChatgptPlugin {
         if (selected) {
             const current = await this.client.activeSession?.().catch(() => undefined);
             const entry = current?.id === selected.id && current.host === selected.host ? current : selected;
-            store.remember(entry, message, reply);
+            store.remember(entry, message, typeof reply === "string" ? reply : reply.text);
             if (request.sessionId)
                 store.bindings[request.sessionId] = entry;
             this.persist();

package/dist/src-ts/plugins/desktop-launch.js

@@ -7,10 +7,19 @@ exports.waitFor = waitFor;
 const promises_1 = require("node:fs/promises");
 const node_fs_1 = require("node:fs");
 const node_path_1 = require("node:path");
+const versionDirectory = /^(?:app[-_ ]?)?v?\d+(?:[.-]\d+)+$/i;
 /** 在 Windows 常见安装目录中查找桌面程序，避免依赖单一盘符。 */
 async function findWindowsExecutable(fileName, configured) {
     if (configured?.trim()) {
         const configuredPath = (0, node_path_1.resolve)(configured);
+        if (fileName.toLowerCase() === "doubao.exe") {
+            let root = (0, node_path_1.dirname)(configuredPath);
+            while (/^app$/i.test((0, node_path_1.basename)(root)) || versionDirectory.test((0, node_path_1.basename)(root)))
+                root = (0, node_path_1.dirname)(root);
+            const latest = await findExecutableInTree(root, fileName, /^/, 3);
+            if (latest)
+                return latest;
+        }
         // 千问升级会保留或删除旧版本目录，从同一安装位置重新定位新版。
         if (fileName.toLowerCase() === "qwenworkcn.exe" && /^v?\d+(?:[.-]\d+)+$/i.test((0, node_path_1.basename)((0, node_path_1.dirname)(configuredPath)))) {
             const latest = await findExecutableInTree((0, node_path_1.dirname)((0, node_path_1.dirname)(configuredPath)), fileName, /^/, 1);
@@ -42,6 +51,16 @@ async function findExecutableInTree(directory, fileName, pattern, depth) {
 async function search(directory, fileName, pattern, depth, insideProduct) {
     try {
         const entries = await (0, promises_1.readdir)(directory, { withFileTypes: true });
+        // 豆包上层文件是启动器，优先查找新版目录或 app 中的实际程序。
+        if (insideProduct && depth > 0 && fileName.toLowerCase() === "doubao.exe") {
+            const versions = entries.filter(entry => entry.isDirectory() && (versionDirectory.test(entry.name) || /^app$/i.test(entry.name)));
+            versions.sort((left, right) => right.name.replace(/^\D+/, "").localeCompare(left.name.replace(/^\D+/, ""), "en", { numeric: true }));
+            for (const entry of versions) {
+                const found = await search((0, node_path_1.join)(directory, entry.name), fileName, pattern, depth - 1, true);
+                if (found)
+                    return found;
+            }
+        }
         for (const entry of entries)
             if (entry.isFile() && insideProduct && entry.name.toLowerCase() === fileName.toLowerCase())
                 return (0, node_path_1.join)(directory, entry.name);

package/dist/src-ts/plugins/doubao/attachments.js

@@ -0,0 +1,42 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.cleanupUploadScript = exports.uploadStateScript = void 0;
+exports.buildUploadInjection = buildUploadInjection;
+// 原生上传状态保存在附件区的 React 属性中，避免复刻平台私有上传签名。
+exports.uploadStateScript = `() => {
+  const area = document.querySelector('[data-testid=attachment_area]');
+  let fiber = area?.[Object.keys(area).find(key => key.startsWith('__reactFiber'))];
+  while (fiber) {
+    if (Array.isArray(fiber.memoizedProps?.attachmentStates)) return fiber.memoizedProps.attachmentStates;
+    fiber = fiber.return;
+  }
+  return [];
+}`;
+function buildUploadInjection(items) {
+    return `(() => {
+    const editor = document.querySelector('[contenteditable=true]');
+    if (!editor) return '未找到豆包输入区';
+    if (editor.textContent.trim() || document.querySelector('[data-testid=attachment_area]')) return '豆包存在未发送草稿，请先处理草稿';
+    const input = document.querySelector('input[type=file][data-testid=upload-file-input]');
+    if (!input) return '未找到豆包原生附件上传入口';
+    const transfer = new DataTransfer();
+    for (const item of ${JSON.stringify(items)}) {
+      const bytes = Uint8Array.from(atob(item.base64), char => char.charCodeAt(0));
+      transfer.items.add(new File([bytes], item.uploadName, { type: item.type }));
+    }
+    Object.defineProperty(input, 'files', { value: transfer.files, configurable: true });
+    try { input.dispatchEvent(new Event('change', { bubbles: true })); }
+    finally { delete input.files; }
+    return 'started';
+  })()`;
+}
+// 唯一文件名前缀限定清理范围，用户上传的新附件不会被删除。
+exports.cleanupUploadScript = `(names) => {
+  const states = (${exports.uploadStateScript})();
+  const area = document.querySelector('[data-testid=attachment_area]');
+  const buttons = area?.querySelectorAll('[data-testid=attachment-delete-btn], [data-testid=message_nested_content_file_delete]') || [];
+  for (let index = states.length - 1; index >= 0; index--) {
+    if (names.includes(states[index].fileName)) buttons[index]?.dispatchEvent(new MouseEvent('click', { bubbles: true }));
+  }
+  return 'cleaned';
+}`;