wechat-to-anything 0.5.1 → 0.5.3

package/README.md

@@ -112,7 +112,9 @@ npx wechat-to-anything \
 
 多 Agent 模式下回复自动带 `[agentName]` 前缀标识来源。每个用户独立维护默认 Agent。
 
-**图片支持**：消息格式遵循 [OpenAI Vision API](https://platform.openai.com/docs/guides/vision)，`content` 为数组：
+## 多媒体格式
+
+**图片（微信 → Agent）**：遵循 [OpenAI Vision API](https://platform.openai.com/docs/guides/vision)，`content` 为数组：
 
 ```json
 {
@@ -126,9 +128,21 @@ npx wechat-to-anything \
 }
 ```
 
-**图片回复**：Agent 回复中包含 markdown 图片 `![desc](https://...)` 会自动作为图片消息发到微信。
+**图片（Agent → 微信）**：回复中包含 `![desc](https://...)` 自动发图。
+
+**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`。
+
+```python
+@app.post("/v1/chat/completions")
+def chat(request):
+    message = request.json["messages"][-1]["content"]
+    audio_path = your_tts(message)  # → /tmp/reply.mp3
+    reply = f"[audio:{audio_path}]\n这是文字版内容"
+    return {"choices": [{"message": {"role": "assistant", "content": reply}}]}
+```
+
+> 完整语音发送示例见 [examples/voice-test.mjs](examples/voice-test.mjs)
 
-**语音回复**：Agent 回复中包含 `[audio:path 或 URL]`，桥会自动转为微信语音消息。支持本地路径和 HTTP URL，格式支持 MP3、WAV、OGG 等。
 ## 凭证
 
 登录凭证保存在 `~/.wechat-to-anything/credentials.json`，删除即可重新登录。

package/cli/bridge.mjs

@@ -320,20 +320,37 @@ export async function start(agents, defaultAgent) {
                 const pcmSize = statSync("/tmp/wxta_audio.pcm").size;
                 const durationMs = Math.round((pcmSize / 32000) * 1000);
 
-                // CDN 上传 + 发送语音
+                // CDN 上传 + 发送语音（与"语音测试"相同格式）
                 const cdn = await uploadToCdn("/tmp/wxta_audio.silk", from, creds.token, 4);
                 const aesKeyB64 = Buffer.from(cdn.aeskey).toString("base64");
-
-                const voiceBody = {
-                  to_user: from,
-                  voice_item: {
-                    voice_url: cdn.file_url, aes_buf_key: aesKeyB64,
-                    file_id: cdn.file_id, voice_length: durationMs, voice_format: 4,
+                const crypto = await import("node:crypto");
+
+                const body = JSON.stringify({
+                  msg: {
+                    from_user_id: "", to_user_id: from,
+                    client_id: crypto.randomUUID(),
+                    message_type: 2, message_state: 2,
+                    item_list: [{
+                      type: 3,
+                      voice_item: {
+                        media: {
+                          encrypt_query_param: cdn.downloadParam,
+                          aes_key: aesKeyB64,
+                        },
+                        encode_type: 4,
+                        bits_per_sample: 16,
+                        sample_rate: 16000,
+                        playtime: durationMs,
+                      },
+                    }],
+                    context_token: contextToken,
                   },
-                };
-                const headers = buildHeaders(creds.token, contextToken);
-                await fetch(`${baseUrl}/cgi-bin/mmchatgpt-wechat/sendvoicemessage`, {
-                  method: "POST", headers, body: JSON.stringify(voiceBody),
+                  base_info: {},
+                });
+                await fetch(`${baseUrl}/ilink/bot/sendmessage`, {
+                  method: "POST",
+                  headers: buildHeaders(creds.token, body),
+                  body,
                 });
                 console.log(pc.green(`→ [语音] 已发送 (${durationMs}ms)`));
                 if (textPart) await sendMessage(creds.token, from, agentTag + textPart, contextToken);

package/cli/weixin.mjs

@@ -187,9 +187,24 @@ export async function sendMessage(token, to, text, contextToken) {
 }
 
 /**
- * 发送图片消息（通过 URL 直接发送）
+ * 发送图片消息（下载 → CDN 上传 → 发送）
  */
 export async function sendImageByUrl(token, to, contextToken, imageUrl) {
+  const { writeFile: wf } = await import("node:fs/promises");
+
+  // 下载图片
+  const resp = await fetch(imageUrl);
+  if (!resp.ok) throw new Error(`图片下载失败: ${resp.status}`);
+  const buf = Buffer.from(await resp.arrayBuffer());
+  const tmpPath = "/tmp/wxta_image_send.jpg";
+  await wf(tmpPath, buf);
+
+  // CDN 上传 (mediaType=1 = IMAGE)
+  const { uploadToCdn } = await import("./cdn.mjs");
+  const cdn = await uploadToCdn(tmpPath, to, token, 1);
+  const aesKeyB64 = Buffer.from(cdn.aeskey).toString("base64");
+
+  // 发送
   await apiPost(
     "ilink/bot/sendmessage",
     {
@@ -202,7 +217,10 @@ export async function sendImageByUrl(token, to, contextToken, imageUrl) {
         item_list: [{
           type: 2, // IMAGE
           image_item: {
-            url: imageUrl,
+            media: {
+              encrypt_query_param: cdn.downloadParam,
+              aes_key: aesKeyB64,
+            },
           },
         }],
         context_token: contextToken,

package/examples/image-test.mjs

@@ -0,0 +1,21 @@
+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/thumb/4/47/PNG_transparency_demonstration_1.png/280px-PNG_transparency_demonstration_1.png";
+
+// 获取 contextToken
+const { getUpdates, buildHeaders, BASE_URL } = await import("../cli/weixin.mjs");
+const msgs = await getUpdates(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");

package/examples/voice-test.mjs

@@ -0,0 +1,62 @@
+import { readFileSync, statSync } from "fs";
+import { homedir } from "os";
+import { execSync } from "child_process";
+import crypto from "crypto";
+
+// 凭证
+const creds = JSON.parse(readFileSync(homedir() + "/.wechat-to-anything/credentials.json", "utf-8"));
+const token = creds.token;
+const to = creds.userId;
+
+// 1. MP3 → PCM → SILK
+console.log("1. 转码 MP3 → PCM → SILK");
+execSync(`ffmpeg -y -i /tmp/wxta_test_voice.mp3 -ar 16000 -ac 1 -f s16le /tmp/wxta_test_voice.pcm 2>/dev/null`);
+execSync(`python3 -c "import pilk; pilk.encode('/tmp/wxta_test_voice.pcm', '/tmp/wxta_test_voice.silk', pcm_rate=16000, tencent=True)"`);
+const pcmSize = statSync("/tmp/wxta_test_voice.pcm").size;
+const durationMs = Math.round((pcmSize / 32000) * 1000);
+console.log(`   SILK ok, duration=${durationMs}ms`);
+
+// 2. CDN 上传
+console.log("2. CDN 上传");
+const { uploadToCdn } = await import("../cli/cdn.mjs");
+const cdn = await uploadToCdn("/tmp/wxta_test_voice.silk", to, token, 4);
+const aesKeyB64 = Buffer.from(cdn.aeskey).toString("base64");
+console.log(`   CDN ok, downloadParam=${cdn.downloadParam.slice(0, 30)}...`);
+
+// 3. 获取 contextToken
+const { getUpdates, buildHeaders, BASE_URL } = await import("../cli/weixin.mjs");
+const msgs = await getUpdates(token);
+const contextToken = msgs?.context_token || "";
+
+// 4. 发送语音（与"语音测试"完全一致的格式）
+console.log("3. 发送语音");
+const body = JSON.stringify({
+  msg: {
+    from_user_id: "", to_user_id: to,
+    client_id: crypto.randomUUID(),
+    message_type: 2, message_state: 2,
+    item_list: [{
+      type: 3,
+      voice_item: {
+        media: {
+          encrypt_query_param: cdn.downloadParam,
+          aes_key: aesKeyB64,
+        },
+        encode_type: 4,
+        bits_per_sample: 16,
+        sample_rate: 16000,
+        playtime: durationMs,
+      },
+    }],
+    context_token: contextToken,
+  },
+  base_info: {},
+});
+
+const res = await fetch(`${BASE_URL}/ilink/bot/sendmessage`, {
+  method: "POST",
+  headers: buildHeaders(token, body),
+  body,
+});
+const text = await res.text();
+console.log(`   结果: ${res.status} ${text}`);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wechat-to-anything",
-  "version": "0.5.1",
+  "version": "0.5.3",
   "description": "一条命令，把微信变成任何 AI Agent 的入口",
   "type": "module",
   "bin": {