feat: image_proxy v2 + stronger SOUL.md

SOUL.md

@@ -28,9 +28,22 @@
 - 对数字化工具（SAP、Oracle、WMS、TMS、APS）信手拈来
 - 敢于挑战传统思维，提出反直觉的洞察
 
-## 图片处理
-当用户发送图片时，消息中会包含 image_key（格式如 `img_v3_xxx`）。你必须立即执行以下命令下载并获取图片 URL：
+## 图片处理（重要！必须执行！）
+当用户发送了图片、截图、照片或转发了包含图片的消息，你必须使用 bash 工具执行以下命令来获取图片：
+```bash
+python3 /app/image_proxy.py --recent
+```
+该脚本会自动查找最近聊天中的图片，下载并上传到图床，返回一个可访问的 URL。
+
+如果你知道具体的 chat_id，也可以用：
+```bash
+python3 /app/image_proxy.py --chat <chat_id>
+```
+
+如果你在消息中看到了 image_key（格式如 img_v3_xxx），直接用：
 ```bash
 python3 /app/image_proxy.py <image_key>
 ```
-拿到 URL 后用 `web_fetch` 获取图片内容进行分析。不要跳过这一步，不要让用户重新发送图片。
+
+拿到 URL 后，你就可以基于图片内容进行分析和回复。
+**绝对不要跳过这一步。绝对不要假装看到了图片。绝对不要让用户重新发送。**

image_proxy.py

@@ -1,105 +1,179 @@
 #!/usr/bin/env python3
 """
-飞书图片代理：下载飞书消息中的图片 → 上传到免费图床 → 返回可访问 URL
-用法: python3 image_proxy.py <image_key>
+飞书图片代理 v2
+用法:
+  python3 image_proxy.py <image_key>          # 直接用 image_key 下载
+  python3 image_proxy.py --chat <chat_id>     # 自动查找该聊天中最近的图片
+  python3 image_proxy.py --recent             # 查找所有聊天中最近收到的图片
 """
-import os, sys, requests
+import os, sys, json, requests, time
+
+FEISHU_BASE = "https://open.feishu.cn/open-apis"
 
 def get_tenant_token():
-    """获取飞书 tenant_access_token"""
     app_id = os.environ.get("FEISHU_APP_ID")
     app_secret = os.environ.get("FEISHU_APP_SECRET")
     if not app_id or not app_secret:
         print("ERROR: FEISHU_APP_ID / FEISHU_APP_SECRET 未设置", file=sys.stderr)
         sys.exit(1)
-    resp = requests.post(
-        "https://open.feishu.cn/open-apis/auth/v3/tenant_access_token/internal",
-        json={"app_id": app_id, "app_secret": app_secret}
-    )
+    resp = requests.post(f"{FEISHU_BASE}/auth/v3/tenant_access_token/internal",
+        json={"app_id": app_id, "app_secret": app_secret})
     data = resp.json()
     if data.get("code") != 0:
         print(f"ERROR: 获取 token 失败: {data}", file=sys.stderr)
         sys.exit(1)
     return data["tenant_access_token"]
 
+def find_image_keys_in_chat(token, chat_id, limit=20):
+    """从指定聊天中查找最近的图片消息"""
+    headers = {"Authorization": f"Bearer {token}"}
+    resp = requests.get(f"{FEISHU_BASE}/im/v1/messages",
+        headers=headers,
+        params={"container_id_type": "chat", "container_id": chat_id,
+                "sort_type": "ByCreateTimeDesc", "page_size": limit})
+    data = resp.json()
+    if data.get("code") != 0:
+        print(f"ERROR: 获取消息列表失败: {data}", file=sys.stderr)
+        return []
+    
+    image_keys = []
+    for item in data.get("data", {}).get("items", []):
+        msg_type = item.get("msg_type", "")
+        body = item.get("body", {}).get("content", "{}")
+        try:
+            content = json.loads(body) if isinstance(body, str) else body
+        except:
+            continue
+        
+        if msg_type == "image":
+            key = content.get("image_key", "")
+            if key:
+                image_keys.append(key)
+        elif msg_type == "post":  # rich text can contain images
+            # Parse post format for image keys
+            for lang_content in (content.get("zh_cn", content.get("en_us", content)) if isinstance(content, dict) else []):
+                if isinstance(lang_content, dict):
+                    for row in lang_content.get("content", []):
+                        if isinstance(row, list):
+                            for elem in row:
+                                if isinstance(elem, dict) and elem.get("tag") == "img":
+                                    key = elem.get("image_key", "")
+                                    if key:
+                                        image_keys.append(key)
+    return image_keys
+
+def find_recent_chats(token):
+    """获取 bot 最近的聊天列表"""
+    headers = {"Authorization": f"Bearer {token}"}
+    resp = requests.get(f"{FEISHU_BASE}/im/v1/chats",
+        headers=headers, params={"page_size": 10, "sort_type": "ByCreateTimeDesc"})
+    data = resp.json()
+    if data.get("code") != 0:
+        print(f"ERROR: 获取聊天列表失败: {data}", file=sys.stderr)
+        return []
+    return [item.get("chat_id") for item in data.get("data", {}).get("items", [])]
+
 def download_image(image_key, token):
-    """从飞书 API 下载图片"""
-    resp = requests.get(
-        f"https://open.feishu.cn/open-apis/im/v1/images/{image_key}",
-        headers={"Authorization": f"Bearer {token}"},
-        stream=True
-    )
+    resp = requests.get(f"{FEISHU_BASE}/im/v1/images/{image_key}",
+        headers={"Authorization": f"Bearer {token}"}, stream=True)
     if resp.status_code != 200:
         print(f"ERROR: 下载图片失败 (HTTP {resp.status_code}): {resp.text[:200]}", file=sys.stderr)
-        sys.exit(1)
+        return None
     return resp.content
 
 def upload_to_catbox(image_data):
-    """上传到 catbox.moe（免费，无需 API Key）"""
-    resp = requests.post(
-        "https://catbox.moe/user/api.php",
-        data={"reqtype": "fileupload"},
-        files={"filedata": ("image.jpg", image_data, "image/jpeg")}
-    )
-    if resp.status_code == 200 and resp.text.startswith("http"):
-        return resp.text.strip()
+    try:
+        resp = requests.post("https://catbox.moe/user/api.php",
+            data={"reqtype": "fileupload"},
+            files={"filedata": ("image.jpg", image_data, "image/jpeg")},
+            timeout=30)
+        if resp.status_code == 200 and resp.text.startswith("http"):
+            return resp.text.strip()
+    except Exception as e:
+        print(f"⚠️ catbox 失败: {e}", file=sys.stderr)
     return None
 
 def upload_to_0x0(image_data):
-    """备用：上传到 0x0.st"""
-    resp = requests.post(
-        "https://0x0.st",
-        files={"file": ("image.jpg", image_data, "image/jpeg")}
-    )
-    if resp.status_code == 200 and resp.text.startswith("http"):
-        return resp.text.strip()
+    try:
+        resp = requests.post("https://0x0.st",
+            files={"file": ("image.jpg", image_data, "image/jpeg")}, timeout=30)
+        if resp.status_code == 200 and resp.text.startswith("http"):
+            return resp.text.strip()
+    except Exception as e:
+        print(f"⚠️ 0x0 失败: {e}", file=sys.stderr)
     return None
 
 def upload_to_tmpfiles(image_data):
-    """备用：上传到 tmpfiles.org"""
-    resp = requests.post(
-        "https://tmpfiles.org/api/v1/upload",
-        files={"file": ("image.jpg", image_data, "image/jpeg")}
-    )
-    if resp.status_code == 200:
-        data = resp.json()
-        url = data.get("data", {}).get("url", "")
-        # tmpfiles.org 返回的 URL 需要加 /dl/ 才能直接访问
+    try:
+        resp = requests.post("https://tmpfiles.org/api/v1/upload",
+            files={"file": ("image.jpg", image_data, "image/jpeg")}, timeout=30)
+        if resp.status_code == 200:
+            url = resp.json().get("data", {}).get("url", "")
+            if url:
+                return url.replace("tmpfiles.org/", "tmpfiles.org/dl/")
+    except Exception as e:
+        print(f"⚠️ tmpfiles 失败: {e}", file=sys.stderr)
+    return None
+
+def upload_image(image_data):
+    for name, fn in [("catbox", upload_to_catbox), ("0x0", upload_to_0x0), ("tmpfiles", upload_to_tmpfiles)]:
+        print(f"📤 上传到 {name}...", file=sys.stderr)
+        url = fn(image_data)
         if url:
-            return url.replace("tmpfiles.org/", "tmpfiles.org/dl/")
+            print(f"✅ 上传成功: {url}", file=sys.stderr)
+            return url
     return None
 
+def process_image_key(image_key, token):
+    print(f"📥 下载飞书图片: {image_key}", file=sys.stderr)
+    data = download_image(image_key, token)
+    if not data:
+        return None
+    print(f"📥 下载成功 ({len(data)} bytes)", file=sys.stderr)
+    url = upload_image(data)
+    if url:
+        print(url)
+        return url
+    # fallback: save locally
+    path = f"/tmp/{image_key}.jpg"
+    with open(path, "wb") as f:
+        f.write(data)
+    print(f"⚠️ 图床失败，本地保存: {path}", file=sys.stderr)
+    print(path)
+    return path
+
 if __name__ == "__main__":
     if len(sys.argv) < 2:
-        print("用法: python3 image_proxy.py <image_key>")
-        print("  从飞书下载图片并上传到免费图床，返回可访问的 URL")
+        print("用法:")
+        print("  python3 image_proxy.py <image_key>")
+        print("  python3 image_proxy.py --chat <chat_id>")
+        print("  python3 image_proxy.py --recent")
         sys.exit(1)
 
-    image_key = sys.argv[1]
-    print(f"📥 正在下载飞书图片: {image_key}", file=sys.stderr)
-    
     token = get_tenant_token()
-    image_data = download_image(image_key, token)
-    print(f"📥 下载成功 ({len(image_data)} bytes)", file=sys.stderr)
-    
-    # 依次尝试多个图床
-    for name, uploader in [("catbox.moe", upload_to_catbox), 
-                            ("0x0.st", upload_to_0x0),
-                            ("tmpfiles.org", upload_to_tmpfiles)]:
-        print(f"📤 正在上传到 {name}...", file=sys.stderr)
-        try:
-            url = uploader(image_data)
-            if url:
-                print(f"✅ 上传成功: {url}", file=sys.stderr)
-                print(url)  # stdout 只输出 URL，供 agent 捕获
+
+    if sys.argv[1] == "--chat" and len(sys.argv) >= 3:
+        chat_id = sys.argv[2]
+        print(f"🔍 在聊天 {chat_id} 中查找图片...", file=sys.stderr)
+        keys = find_image_keys_in_chat(token, chat_id)
+        if not keys:
+            print("❌ 未找到图片", file=sys.stderr)
+            sys.exit(1)
+        print(f"🔍 找到 {len(keys)} 张图片，处理第一张: {keys[0]}", file=sys.stderr)
+        process_image_key(keys[0], token)
+
+    elif sys.argv[1] == "--recent":
+        print("🔍 查找最近聊天中的图片...", file=sys.stderr)
+        chats = find_recent_chats(token)
+        for chat_id in chats:
+            keys = find_image_keys_in_chat(token, chat_id, limit=5)
+            if keys:
+                print(f"🔍 在 {chat_id} 找到图片: {keys[0]}", file=sys.stderr)
+                process_image_key(keys[0], token)
                 sys.exit(0)
-        except Exception as e:
-            print(f"⚠️ {name} 上传失败: {e}", file=sys.stderr)
-    
-    # 所有图床都失败，保存到本地
-    local_path = f"/tmp/{image_key}.jpg"
-    with open(local_path, "wb") as f:
-        f.write(image_data)
-    print(f"⚠️ 所有图床失败，已保存到本地: {local_path}", file=sys.stderr)
-    print(local_path)
-    sys.exit(1)
+        print("❌ 所有聊天中均未找到图片", file=sys.stderr)
+        sys.exit(1)
+
+    else:
+        image_key = sys.argv[1]
+        process_image_key(image_key, token)