wechat-to-anything 0.5.6 → 0.6.0

package/README.md

@@ -128,7 +128,7 @@ npx wechat-to-anything \
 }
 ```
 
-**图片（Agent → 微信）**：回复中包含 `![desc](https://...)` 自动发图。
+**图片（Agent → 微信）**：回复中包含 `![desc](https://...)` 自动发图（HD 原图质量，自动生成缩略图）。支持 URL、本地路径、data URI。
 
 **语音（Agent → 微信）**：回复中包含 `[audio:path 或 URL]` 自动发语音气泡。支持 MP3、WAV、OGG 等。需要 `ffmpeg` 和 `pip install pilk`。
 

package/cli/bridge.mjs

@@ -1,11 +1,14 @@
 import pc from "picocolors";
 import {
   loadCredentials, loginWithQR, getUpdates,
-  sendMessage, sendImageByUrl,
+  sendMessage, sendImageByUrl, sendVideoByUrl,
   extractText, extractMedia,
+  getConfig, sendTyping,
 } from "./weixin.mjs";
-import { downloadAndDecrypt, downloadMediaToFile } from "./cdn.mjs";
+import { downloadAndDecrypt, downloadMediaToFile, uploadToCdn } from "./cdn.mjs";
 import { callAgentAuto, checkAgent } from "./agent-adapter.mjs";
+import { createTypingController } from "./typing.mjs";
+import { stripMarkdown } from "./markdown.mjs";
 
 /**
  * 启动桥：WeChat ilinkai API ←→ Agent HTTP
@@ -74,6 +77,28 @@ export async function start(agents, defaultAgent) {
 
   let getUpdatesBuf = "";
 
+  // typing_ticket 缓存（参考 openclaw-weixin config-cache.ts: 24h TTL）
+  const typingTickets = new Map(); // userId → { ticket, fetchedAt }
+  const TICKET_TTL = 24 * 60 * 60_000; // 24h
+
+  async function getTypingTicket(userId, contextToken) {
+    const cached = typingTickets.get(userId);
+    if (cached && (Date.now() - cached.fetchedAt) < TICKET_TTL) {
+      return cached.ticket;
+    }
+    try {
+      const cfg = await getConfig(creds.token, userId, contextToken);
+      const ticket = cfg?.typing_ticket || "";
+      if (ticket) {
+        typingTickets.set(userId, { ticket, fetchedAt: Date.now() });
+        console.log(pc.dim(`   typing_ticket cached for ${userId.slice(0, 8)}...`));
+      }
+      return ticket;
+    } catch {
+      return cached?.ticket || "";
+    }
+  }
+
   // 每用户图片缓存：发图片时先存着，等下一条文字消息再合并发出
   const pendingImages = new Map(); // userId → { base64, timestamp }
   const IMAGE_BUFFER_TTL = 5 * 60_000; // 5 min 过期
@@ -285,15 +310,33 @@ export async function start(agents, defaultAgent) {
           }
           const agentUrl = agents.get(targetAgent);
 
+          // Typing 指示器：Agent 思考期间显示"正在输入..."
+          const typingTicket = await getTypingTicket(from, contextToken);
+          const typing = typingTicket
+            ? createTypingController({
+                start: () => sendTyping(creds.token, from, typingTicket, 1),
+                stop:  () => sendTyping(creds.token, from, typingTicket, 2),
+                onError: (err) => console.error(pc.dim(`   typing: ${err.message}`)),
+                keepaliveMs: 5000,
+                maxDurationMs: 60_000,
+              })
+            : null;
+
           // 调用 Agent
           try {
+            if (typing) await typing.onReplyStart();
             const reply = await callAgentAuto(agentUrl, agentMessages);
+            if (typing) typing.onIdle();
             const agentTag = multiMode ? `[${targetAgent}] ` : "";
 
             // 检查回复是否包含 [audio:path/url]
             const audioMatch = reply.match(/\[audio:(.*?)\]/);
             // 检查回复是否包含图片（markdown 格式，支持 URL 和 data URI）
             const imageMatch = reply.match(/!\[.*?\]\(((?:https?:\/\/|data:image\/)[^\s)]+)\)/);
+            // 检查回复是否包含 [video:path/url]
+            const videoMatch = reply.match(/\[video:(.*?)\]/);
+            // 检查回复是否包含 [file:path/url]
+            const fileMatch = reply.match(/\[file:(.*?)\]/);
 
             if (audioMatch) {
               const audioSrc = audioMatch[1];
@@ -353,7 +396,7 @@ export async function start(agents, defaultAgent) {
                   body,
                 });
                 console.log(pc.green(`→ [语音] 已发送 (${durationMs}ms)`));
-                if (textPart) await sendMessage(creds.token, from, agentTag + textPart, contextToken);
+                if (textPart) await sendMessage(creds.token, from, agentTag + stripMarkdown(textPart), contextToken);
               } catch (err) {
                 console.error(pc.red(`   语音发送失败: ${err.message}`));
                 await sendMessage(creds.token, from, agentTag + reply.replace(/\[audio:.*?\]/g, "").trim() || reply, contextToken);
@@ -364,18 +407,58 @@ export async function start(agents, defaultAgent) {
               const textPart = reply.replace(/!\[.*?\]\(https?:\/\/[^\s)]+\)/g, "").trim();
               console.log(pc.green(`→ [${targetAgent}] [图片] ${imageUrl.slice(0, 60)}`));
               try {
-                if (textPart) await sendMessage(creds.token, from, agentTag + textPart, contextToken);
+                if (textPart) await sendMessage(creds.token, from, agentTag + stripMarkdown(textPart), contextToken);
                 await sendImageByUrl(creds.token, from, contextToken, imageUrl);
               } catch (err) {
                 console.error(pc.red(`   图片发送失败: ${err.message}`));
                 await sendMessage(creds.token, from, agentTag + reply, contextToken);
               }
+            } else if (videoMatch) {
+              // Agent 回复了视频 → CDN 上传发到微信
+              const videoSrc = videoMatch[1];
+              const textPart = reply.replace(/\[video:.*?\]/g, "").trim();
+              console.log(pc.green(`→ [${targetAgent}] [视频] ${videoSrc.slice(0, 60)}`));
+              try {
+                if (textPart) await sendMessage(creds.token, from, agentTag + stripMarkdown(textPart), contextToken);
+                await sendVideoByUrl(creds.token, from, contextToken, videoSrc);
+              } catch (err) {
+                console.error(pc.red(`   视频发送失败: ${err.message}`));
+                await sendMessage(creds.token, from, agentTag + stripMarkdown(reply), contextToken);
+              }
+            } else if (fileMatch) {
+              // Agent 回复了文件 → CDN 上传发到微信
+              const fileSrc = fileMatch[1];
+              const textPart = reply.replace(/\[file:.*?\]/g, "").trim();
+              const fileName = fileSrc.split("/").pop() || "file";
+              console.log(pc.green(`→ [${targetAgent}] [文件] ${fileSrc.slice(0, 60)}`));
+              try {
+                const { writeFileSync, unlinkSync } = await import("node:fs");
+                const { tmpdir } = await import("node:os");
+                const { join } = await import("node:path");
+                const resp = await fetch(fileSrc);
+                if (!resp.ok) throw new Error(`file download failed: ${resp.status}`);
+                const buf = Buffer.from(await resp.arrayBuffer());
+                const tmpPath = join(tmpdir(), `wx-file-${Date.now()}-${fileName}`);
+                writeFileSync(tmpPath, buf);
+                try {
+                  const uploaded = await uploadToCdn(tmpPath, from, creds.token, 3);
+                  const { sendFileMessage } = await import("./weixin.mjs");
+                  await sendFileMessage(creds.token, from, contextToken, uploaded, fileName);
+                  if (textPart) await sendMessage(creds.token, from, agentTag + stripMarkdown(textPart), contextToken);
+                } finally {
+                  try { unlinkSync(tmpPath); } catch {}
+                }
+              } catch (err) {
+                console.error(pc.red(`   文件发送失败: ${err.message}`));
+                await sendMessage(creds.token, from, agentTag + stripMarkdown(reply), contextToken);
+              }
             } else {
               // 纯文本回复
               console.log(pc.green(`→ [${targetAgent}] ${reply.slice(0, 80)}${reply.length > 80 ? "..." : ""}`));
-              await sendMessage(creds.token, from, agentTag + reply, contextToken);
+              await sendMessage(creds.token, from, agentTag + stripMarkdown(reply), contextToken);
             }
           } catch (err) {
+            if (typing) typing.onCleanup();
             console.error(pc.red(`   ${targetAgent} 错误: ${err.message}`));
             await sendMessage(creds.token, from, `⚠️ ${targetAgent} 错误: ${err.message}`, contextToken);
           }

package/cli/markdown.mjs

@@ -0,0 +1,53 @@
+/**
+ * Markdown → 纯文本
+ *
+ * 微信不渲染 markdown，需要剥离语法。
+ *
+ * 2 层剥离，参考 OpenClaw：
+ *   层 1: openclaw-weixin send.ts#L20-35 — 代码块/图片/链接/表格
+ *   层 2: OpenClaw markdown-to-line.ts#L344-375 — bold/italic/headers/quotes/rules
+ */
+
+/**
+ * 剥离 markdown 语法，返回纯文本
+ * @param {string} text
+ * @returns {string}
+ */
+export function stripMarkdown(text) {
+  let r = text;
+
+  // === 层 1: 结构化 markdown（weixin send.ts） ===
+  // 代码块 → 保留内容
+  r = r.replace(/```[^\n]*\n?([\s\S]*?)```/g, (_, code) => code.trim());
+  // 图片标记 → 删除
+  r = r.replace(/!\[[^\]]*\]\([^)]*\)/g, "");
+  // 链接 → 保留文字
+  r = r.replace(/\[([^\]]+)\]\([^)]*\)/g, "$1");
+  // 表格分隔行 → 删除
+  r = r.replace(/^\|[\s:|-]+\|$/gm, "");
+  // 表格行 → 空格分隔
+  r = r.replace(/^\|(.+)\|$/gm, (_, inner) =>
+    inner.split("|").map((c) => c.trim()).join("  "));
+
+  // === 层 2: inline markdown（markdown-to-line.ts） ===
+  // **粗体** / __粗体__
+  r = r.replace(/\*\*(.+?)\*\*/g, "$1");
+  r = r.replace(/__(.+?)__/g, "$1");
+  // *斜体*（lookbehind 避免误匹配 **）
+  r = r.replace(/(?<!\*)\*(?!\*)(.+?)(?<!\*)\*(?!\*)/g, "$1");
+  r = r.replace(/(?<!_)_(?!_)(.+?)(?<!_)_(?!_)/g, "$1");
+  // ~~删除线~~
+  r = r.replace(/~~(.+?)~~/g, "$1");
+  // # 标题 → 保留文字
+  r = r.replace(/^#{1,6}\s+(.+)$/gm, "$1");
+  // > 引用 → 保留内容
+  r = r.replace(/^>\s?(.*)$/gm, "$1");
+  // ---, ***, ___ 分隔线 → 删除
+  r = r.replace(/^[-*_]{3,}$/gm, "");
+  // `行内代码` → 保留内容
+  r = r.replace(/`([^`]+)`/g, "$1");
+  // 多余空行
+  r = r.replace(/\n{3,}/g, "\n\n");
+
+  return r.trim();
+}

package/cli/typing.mjs

@@ -0,0 +1,189 @@
+/**
+ * Typing 指示器控制器
+ *
+ * 学习 OpenClaw 框架的 3 层架构：
+ *   1. StartGuard  — 熔断器，连续失败 N 次停止重试
+ *   2. KeepaliveLoop — 定时续发 typing（微信 ~10s 自动取消）
+ *   3. TypingController — 组合上述两层 + TTL 安全阀
+ *
+ * 参考:
+ *   OpenClaw/src/channels/typing-start-guard.ts
+ *   OpenClaw/src/channels/typing-lifecycle.ts
+ *   OpenClaw/src/channels/typing.ts
+ *
+ * 用法:
+ *   const ctrl = createTypingController({ start, stop, onError });
+ *   await ctrl.onReplyStart();  // 收到消息，准备调 Agent 前
+ *   const reply = await callAgent();
+ *   ctrl.onIdle();              // Agent 完成，发消息前
+ */
+
+// ═══════════════════════════════════════════════════════════
+// 层 1: StartGuard — 熔断器
+// 参考: OpenClaw/src/channels/typing-start-guard.ts
+// ═══════════════════════════════════════════════════════════
+
+/**
+ * 连续失败 maxFailures 次后熔断（tripped），停止重试。
+ * 错误不对外 throw，只回调 onError → 不影响主流程。
+ */
+function createStartGuard({ onError, maxFailures = 2 } = {}) {
+  let consecutiveFailures = 0;
+  let tripped = false;
+
+  return {
+    async run(fn) {
+      if (tripped) return "skipped";
+      try {
+        await fn();
+        consecutiveFailures = 0;
+        return "started";
+      } catch (err) {
+        consecutiveFailures += 1;
+        onError?.(err);
+        if (consecutiveFailures >= maxFailures) {
+          tripped = true;
+          return "tripped";
+        }
+        return "failed";
+      }
+    },
+    reset() {
+      consecutiveFailures = 0;
+      tripped = false;
+    },
+    isTripped() {
+      return tripped;
+    },
+  };
+}
+
+// ═══════════════════════════════════════════════════════════
+// 层 2: KeepaliveLoop — 心跳循环
+// 参考: OpenClaw/src/channels/typing-lifecycle.ts
+// ═══════════════════════════════════════════════════════════
+
+/**
+ * 每 intervalMs 毫秒执行一次 onTick。
+ * tickInFlight 锁防止重入（上一次未完成不发新请求）。
+ */
+function createKeepaliveLoop({ intervalMs, onTick }) {
+  let timer = undefined;
+  let tickInFlight = false;
+
+  const tick = async () => {
+    if (tickInFlight) return;
+    tickInFlight = true;
+    try {
+      await onTick();
+    } finally {
+      tickInFlight = false;
+    }
+  };
+
+  return {
+    start() {
+      if (intervalMs <= 0 || timer) return;
+      timer = setInterval(() => { void tick(); }, intervalMs);
+    },
+    stop() {
+      if (!timer) return;
+      clearInterval(timer);
+      timer = undefined;
+      tickInFlight = false;
+    },
+    isRunning() {
+      return timer !== undefined;
+    },
+  };
+}
+
+// ═══════════════════════════════════════════════════════════
+// 层 3: TypingController — 对外接口
+// 参考: OpenClaw/src/channels/typing.ts
+// ═══════════════════════════════════════════════════════════
+
+/**
+ * 创建 typing 控制器。
+ *
+ * @param {Object} params
+ * @param {() => Promise<void>} params.start   — 发 sendTyping(TYPING)
+ * @param {() => Promise<void>} [params.stop]  — 发 sendTyping(CANCEL)
+ * @param {(err: Error) => void} params.onError — 错误回调（不影响主流程）
+ * @param {number} [params.keepaliveMs=5000]   — 心跳间隔
+ * @param {number} [params.maxDurationMs=60000] — TTL 安全阀
+ * @param {number} [params.maxFailures=2]       — 熔断阈值
+ * @returns {{ onReplyStart: () => Promise<void>, onIdle: () => void, onCleanup: () => void }}
+ */
+export function createTypingController({
+  start,
+  stop,
+  onError,
+  keepaliveMs = 5000,
+  maxDurationMs = 60_000,
+  maxFailures = 2,
+} = {}) {
+  let stopSent = false;
+  let closed = false;
+  let ttlTimer = undefined;
+
+  const guard = createStartGuard({ onError, maxFailures });
+
+  const fireStart = async () => {
+    const result = await guard.run(() => start());
+    if (result === "tripped") {
+      keepalive.stop();
+    }
+    return result;
+  };
+
+  const keepalive = createKeepaliveLoop({
+    intervalMs: keepaliveMs,
+    onTick: fireStart,
+  });
+
+  // TTL 安全阀：超过 maxDurationMs 自动停止
+  const startTtlTimer = () => {
+    if (maxDurationMs <= 0) return;
+    clearTtlTimer();
+    ttlTimer = setTimeout(() => {
+      if (!closed) fireStop();
+    }, maxDurationMs);
+  };
+
+  const clearTtlTimer = () => {
+    if (ttlTimer) {
+      clearTimeout(ttlTimer);
+      ttlTimer = undefined;
+    }
+  };
+
+  // 开始 typing（调 Agent 前调用）
+  const onReplyStart = async () => {
+    if (closed) return;
+    stopSent = false;
+    guard.reset();
+    keepalive.stop();
+    clearTtlTimer();
+    await fireStart();
+    if (guard.isTripped()) return;
+    keepalive.start();
+    startTtlTimer();
+  };
+
+  // 停止 typing（Agent 完成后调用）
+  const fireStop = () => {
+    closed = true;
+    keepalive.stop();
+    clearTtlTimer();
+    if (!stop || stopSent) return;
+    stopSent = true;
+    void stop().catch((err) => onError?.(err));
+  };
+
+  return {
+    onReplyStart,
+    onIdle: fireStop,
+    onCleanup: fireStop,
+  };
+}

package/cli/weixin.mjs

@@ -10,9 +10,9 @@
  */
 
 import crypto from "node:crypto";
-import { readFileSync, writeFileSync, mkdirSync, existsSync } from "node:fs";
-import { resolve } from "node:path";
-import { homedir } from "node:os";
+import { readFileSync, writeFileSync, mkdirSync, existsSync, unlinkSync } from "node:fs";
+import { resolve, join } from "node:path";
+import { homedir, tmpdir } from "node:os";
 
 export const BASE_URL = "https://ilinkai.weixin.qq.com";
 const LONG_POLL_TIMEOUT_MS = 35_000;
@@ -318,6 +318,83 @@ export async function sendFileMessage(token, to, contextToken, uploaded, fileNam
   );
 }
 
+/**
+ * 发送视频消息（通过 URL 下载 → CDN 上传 → 发送）
+ * 参考: openclaw-weixin send.ts#L209-233 (sendVideoMessageWeixin)
+ */
+export async function sendVideoByUrl(token, to, contextToken, videoUrl) {
+  const { uploadToCdn } = await import("./cdn.mjs");
+
+  // 下载视频
+  const resp = await fetch(videoUrl);
+  if (!resp.ok) throw new Error(`video download failed: ${resp.status}`);
+  const buf = Buffer.from(await resp.arrayBuffer());
+  const tmpPath = join(tmpdir(), `wx-video-${Date.now()}.mp4`);
+  writeFileSync(tmpPath, buf);
+
+  try {
+    // CDN 上传 (mediaType=2=VIDEO)
+    const uploaded = await uploadToCdn(tmpPath, to, token, 2);
+
+    // 发送 type:5 video_item
+    await apiPost(
+      "ilink/bot/sendmessage",
+      {
+        msg: {
+          from_user_id: "",
+          to_user_id: to,
+          client_id: crypto.randomUUID(),
+          message_type: 2,
+          message_state: 2,
+          item_list: [{
+            type: 5, // VIDEO
+            video_item: {
+              media: {
+                encrypt_query_param: uploaded.downloadParam,
+                aes_key: Buffer.from(uploaded.aeskey, "hex").toString("base64"),
+              },
+              video_size: uploaded.ciphertextSize,
+            },
+          }],
+          context_token: contextToken,
+        },
+        base_info: {},
+      },
+      token,
+      API_TIMEOUT_MS
+    );
+  } finally {
+    try { unlinkSync(tmpPath); } catch {}
+  }
+}
+
+/**
+ * 获取 bot 配置（含 typing_ticket）
+ * 参考: openclaw-weixin api.ts#L209-226
+ */
+export async function getConfig(token, userId, contextToken) {
+  return apiPost(
+    "ilink/bot/getconfig",
+    { ilink_user_id: userId, context_token: contextToken },
+    token,
+    10_000
+  );
+}
+
+/**
+ * 发送打字指示器
+ * 参考: openclaw-weixin api.ts#L228-240
+ * @param {number} status — 1=typing, 2=cancel
+ */
+export async function sendTyping(token, userId, typingTicket, status = 1) {
+  return apiPost(
+    "ilink/bot/sendtyping",
+    { ilink_user_id: userId, typing_ticket: typingTicket, status },
+    token,
+    10_000
+  );
+}
+
 /**
  * 提取消息文本（支持语音转文字）
  */

package/examples/image-test.mjs

@@ -1,21 +1,45 @@
+/**
+ * 图片发送测试 — HD 原图质量，自动生成缩略图
+ *
+ * 支持三种输入：
+ *   - URL:       https://example.com/photo.jpg
+ *   - 本地路径:   /path/to/photo.jpg
+ *   - data URI:  data:image/jpeg;base64,...
+ *
+ * 用法:  node examples/image-test.mjs [图片URL或路径]
+ *
+ * ── 关键参数说明 ──
+ *
+ * CDN 上传阶段 (getUploadUrl):
+ *   no_need_thumb: false    — 必须为 false，否则微信只显示模糊压缩图
+ *   thumb_rawsize           — 缩略图明文大小
+ *   thumb_rawfilemd5        — 缩略图明文 MD5
+ *   thumb_filesize          — 缩略图密文大小 (AES-128-ECB padding 后)
+ *
+ * 发送消息阶段 (sendmessage image_item):
+ *   media.encrypt_query_param  — 原图 CDN 下载参数
+ *   media.aes_key              — base64(hex_aeskey) 解密用
+ *   thumb_media                — 缩略图 CDN 引用（没有则显示模糊/不显示）
+ *   mid_size                   — 原图密文大小，微信用于预加载
+ *   thumb_size                 — 缩略图密文大小
+ *   thumb_width                — 缩略图宽度 px，用于聊天列表预留正确尺寸
+ *   thumb_height               — 缩略图高度 px，同上
+ *
+ *   ⚠️  hd_size     — 设置后"查看原图"下载卡 0%，不要用
+ *   ⚠️  encrypt_type — 设置为 1 会导致图片不显示，不要用
+ */
 import { readFileSync } from "fs";
 import { homedir } from "os";
 
-// 凭证
 const creds = JSON.parse(readFileSync(homedir() + "/.wechat-to-anything/credentials.json", "utf-8"));
-const token = creds.token;
-const to = creds.userId;
 
-// 测试图片 URL（可替换为任意图片地址）
-const imageUrl = "https://upload.wikimedia.org/wikipedia/commons/4/47/PNG_transparency_demonstration_1.png";
+// 命令行参数 或 默认测试图片
+const imageUrl = process.argv[2] || "https://upload.wikimedia.org/wikipedia/commons/4/47/PNG_transparency_demonstration_1.png";
 
-// 获取 contextToken
-const { getUpdates, buildHeaders, BASE_URL } = await import("../cli/weixin.mjs");
-const msgs = await getUpdates(token);
+const { sendImageByUrl, getUpdates } = await import("../cli/weixin.mjs");
+const msgs = await getUpdates(creds.token);
 const contextToken = msgs?.context_token || "";
 
-// 发送图片
-console.log("发送图片:", imageUrl.slice(0, 60) + "...");
-const { sendImageByUrl } = await import("../cli/weixin.mjs");
-await sendImageByUrl(token, to, contextToken, imageUrl);
-console.log("✅ 图片已发送");
+console.log("发送图片 (HD):", imageUrl.slice(0, 60) + (imageUrl.length > 60 ? "..." : ""));
+await sendImageByUrl(creds.token, creds.userId, contextToken, imageUrl);
+console.log("✅ 图片已发送（含缩略图，原图质量）");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wechat-to-anything",
-  "version": "0.5.6",
+  "version": "0.6.0",
   "description": "一条命令，把微信变成任何 AI Agent 的入口",
   "type": "module",
   "bin": {