尝试dify读取图片！

plugins/dify/main.py

@@ -5,6 +5,9 @@ import json
 import time
 import re  # 添加re模块导入
 import asyncio
+import base64
+import html
+import xml.etree.ElementTree as ET
 from typing import Dict, Any, List, Optional, Tuple
 
 from pathlib import Path
@@ -20,6 +23,7 @@ from utils.decorator.points_decorator import plugin_points_cost
 from utils.media_downloader import MediaDownloader
 from utils.string_utils import remove_trailing_content, remove_grok_render_tags
 from wechat_ipad import WechatAPIClient
+from wechat_ipad.models.message import MessageType
 import aiohttp
 
 # 常见的图片和视频文件扩展名
@@ -148,6 +152,8 @@ class DifyPlugin(MessagePluginInterface):
 
         # 获取目标接收者（群聊为群ID，私聊为个人wxid）
         target = roomid if roomid else sender
+        self.LOG.debug(
+            f"消息上下文: sender={sender}, roomid={roomid}, target={target}, is_at={message.get('is_at', False)}")
 
         # 检查权限
         if roomid and gbm.get_group_permission(target, self.feature) == PermissionStatus.DISABLED:
@@ -158,7 +164,7 @@ class DifyPlugin(MessagePluginInterface):
             # 去除@的人和空格等字符
             query = re.sub(r"@.*?[\u2005|\s]", "", content).strip()
             if not query:
-                await bot.send_at_message(target, "请在@我的同时提供问题内容", [sender])
+                # await bot.send_at_message(target, "请在@我的同时提供问题内容", [sender])
                 return False, "没有提供问题内容"
         else:
             # 处理命令消息
@@ -168,16 +174,54 @@ class DifyPlugin(MessagePluginInterface):
                 return False, "命令格式错误"
             query = parts[1].strip()
 
+        self.LOG.debug(f"解析请求: query_len={len(query)} query_preview={query[:120]}")
+
         try:
             # 发送等待消息
             # client_msg_id, create_time, new_msg_id = await bot.send_text_message(
             #     target, "⏳AI 正在加油，请稍候… 😊", sender if roomid else "")
             # revoke.add_message_to_revoke(target, client_msg_id, create_time, new_msg_id, 3)
 
-            # TODO，支持@ 消息（比如：图片引用，文本聊天引用，视频引用）
+            dify_files = []
-            # 需要根据引用的信息，进行相关资源下载。如果是表情，则不理解。
+            quote_payload = self._parse_quote_payload(message.get("full_wx_msg"))
-            # 调用Dify API获取回复
+            if quote_payload:
-            success, response = await self._chat_with_dify(target, sender, query)
+                ref_type = quote_payload.get("ref_type", 0)
+                ref_content = quote_payload.get("ref_content", "").strip()
+                title = quote_payload.get("title", "").strip()
+                self.LOG.debug(
+                    f"检测到引用消息: type={ref_type}, title_preview={title[:80]}, ref_len={len(ref_content)}")
+
+                if ref_type == MessageType.TEXT.value:
+                    if ref_content:
+                        self.LOG.debug("使用引用文本作为问题")
+                        query = ref_content
+                elif ref_type == MessageType.IMAGE.value:
+                    if title:
+                        title_query = re.sub(r"@.*?[\u2005|\s]", "", title).strip()
+                        if title_query:
+                            query = title_query
+                    self.LOG.debug("检测到引用图片，开始下载与上传")
+
+                    image_path = await self._download_quote_image(bot, ref_content)
+                    if not image_path:
+                        return False, "图片引用解析失败"
+
+                    upload_id = await self._upload_file_to_dify(image_path, sender)
+                    if not upload_id:
+                        return False, "图片上传失败"
+                    self.LOG.debug(f"图片上传完成: upload_id={upload_id}")
+
+                    dify_files.append({
+                        "type": "image",
+                        "transfer_method": "local_file",
+                        "upload_file_id": upload_id
+                    })
+                elif ref_type == MessageType.VIDEO.value:
+                    return False, "暂不支持视频引用"
+                elif ref_type in (MessageType.EMOTICON.value, MessageType.EMOJI.value):
+                    return False, "暂不支持表情引用"
+
+            success, response = await self._chat_with_dify(target, sender, query, dify_files)
             if not success:
                 return False, response
 
@@ -232,7 +276,147 @@ class DifyPlugin(MessagePluginInterface):
             self.LOG.error(f"发送响应消息时出错: {e}")
             return False, f"发送响应失败: {e}"
 
-    async def _chat_with_dify(self, session_id: str, user_id: str, query: str) -> Tuple[bool, Optional[str]]:
+    def _parse_quote_payload(self, full_msg: Any) -> Optional[Dict[str, Any]]:
+        if not full_msg or not getattr(full_msg, "content", None):
+            return None
+
+        xml_content = getattr(full_msg.content, "xml_content", "")
+        if not xml_content:
+            return None
+
+        try:
+            root = ET.fromstring(xml_content)
+        except ET.ParseError:
+            return None
+
+        appmsg = root.find(".//appmsg")
+        if appmsg is None:
+            return None
+
+        if appmsg.findtext("type", "").strip() != "57":
+            return None
+
+        title = appmsg.findtext("title", "") or ""
+        refer = appmsg.find("refermsg")
+        if refer is None:
+            return None
+
+        ref_type = int(refer.findtext("type", "0") or 0)
+        ref_content = refer.findtext("content", "") or ""
+
+        self.LOG.debug(
+            f"引用解析成功: type={ref_type}, title_len={len(title)}, content_len={len(ref_content)}")
+
+        return {
+            "title": html.unescape(title),
+            "ref_type": ref_type,
+            "ref_content": html.unescape(ref_content)
+        }
+
+    def _extract_quote_image_info(self, ref_content: str) -> Optional[Dict[str, str]]:
+        if not ref_content:
+            return None
+
+        aeskey_match = re.search(r'aeskey="([^"]+)"', ref_content)
+        if not aeskey_match:
+            return None
+
+        url_match = re.search(r'cdnmidimgurl="([^"]+)"', ref_content)
+        if not url_match:
+            url_match = re.search(r'cdnbigimgurl="([^"]+)"', ref_content)
+        if not url_match:
+            url_match = re.search(r'cdnthumburl="([^"]+)"', ref_content)
+
+        if not url_match:
+            return None
+
+        md5_match = re.search(r'md5="([^"]+)"', ref_content)
+
+        return {
+            "aeskey": aeskey_match.group(1),
+            "url": url_match.group(1),
+            "md5": md5_match.group(1) if md5_match else ""
+        }
+
+    async def _download_quote_image(self, bot: WechatAPIClient, ref_content: str) -> Optional[str]:
+        image_info = self._extract_quote_image_info(ref_content)
+        if not image_info:
+            return None
+        self.LOG.debug(
+            f"准备下载引用图片: url_len={len(image_info['url'])}, aeskey_prefix={image_info['aeskey'][:6]}")
+
+        try:
+            base64_str = await bot.download_image(
+                aeskey=image_info["aeskey"],
+                cdnmidimgurl=image_info["url"]
+            )
+        except Exception as e:
+            self.LOG.error(f"下载引用图片失败: {e}")
+            return None
+
+        if not base64_str:
+            return None
+
+        try:
+            image_data = base64.b64decode(base64_str)
+        except Exception as e:
+            self.LOG.error(f"解码引用图片失败: {e}")
+            return None
+
+        temp_dir = Path(__file__).resolve().parents[2] / "temp"
+        os.makedirs(temp_dir, exist_ok=True)
+
+        suffix = image_info["md5"] if image_info.get("md5") else str(int(time.time()))
+        file_path = temp_dir / f"dify_quote_{suffix}.jpg"
+
+        try:
+            with open(file_path, "wb") as f:
+                f.write(image_data)
+        except Exception as e:
+            self.LOG.error(f"保存引用图片失败: {e}")
+            return None
+
+        self.LOG.debug(f"引用图片已保存: {file_path}")
+        return str(file_path)
+
+    async def _upload_file_to_dify(self, file_path: str, user_id: str) -> Optional[str]:
+        if not file_path or not os.path.isfile(file_path):
+            return None
+
+        self.LOG.debug(f"开始上传文件到Dify: {file_path}")
+        url = f"{self.base_url}/files/upload"
+        headers = {"Authorization": f"Bearer {self.api_key}"}
+        proxy = self.http_proxy if self.http_proxy else None
+
+        try:
+            async with aiohttp.ClientSession() as session:
+                with open(file_path, "rb") as f:
+                    form = aiohttp.FormData()
+                    form.add_field("file", f, filename=os.path.basename(file_path))
+                    form.add_field("user", user_id)
+                    response = await session.post(url, headers=headers, data=form, proxy=proxy, timeout=40)
+                    if response.status != 200:
+                        error_text = await response.text()
+                        self.LOG.error(f"Dify上传失败: {response.status} {error_text}")
+                        return None
+
+                    resp_data = await response.json()
+
+            if isinstance(resp_data, dict):
+                if resp_data.get("id"):
+                    return resp_data.get("id")
+                data = resp_data.get("data", {})
+                if isinstance(data, dict):
+                    return data.get("id") or data.get("file_id")
+
+            return None
+
+        except Exception as e:
+            self.LOG.error(f"上传文件到Dify失败: {e}")
+            return None
+
+    async def _chat_with_dify(self, session_id: str, user_id: str, query: str,
+                              files: Optional[List[Dict[str, Any]]] = None) -> Tuple[bool, Optional[str]]:
         """
         与Dify API交互获取回复
         
@@ -281,9 +465,13 @@ class DifyPlugin(MessagePluginInterface):
         if history_text:
             inputs_params["history"] = history_text
 
+        if files is None:
+            files = []
+        self.LOG.debug(f"Dify请求准备: files={len(files)}")
+
         # 准备请求数据
         data = {
-            "sys.files": [],
+            "sys.files": files,
             "user": user_id,
             "inputs": inputs_params,
             "response_mode": "blocking"  # 使用阻塞响应模式