feat: Vision API 图片识别，大师看图说法

image_daemon.py

@@ -4,7 +4,7 @@
 通过 lark-oapi SDK 的 WebSocket 长连接实时接收消息事件，
 检测图片消息后下载、上传到图床、回复 URL。
 """
-import os, sys, json, time, requests, threading
+import os, sys, json, time, requests, threading, base64
 
 FEISHU_BASE = "https://open.feishu.cn/open-apis"
 APP_ID = os.environ.get("FEISHU_APP_ID", "")
@@ -155,9 +155,47 @@ def send_text(token, chat_id, text):
         log(f"❌ 发送消息失败 (code={code}): {data.get('msg', '')}")
     return data
 
+# ---------- Vision 图片分析 ----------
+def analyze_image_with_vision(img_data):
+    """将图片 base64 传给 LLM，由大师以人行佛教视角描述"""
+    if not API_KEY:
+        return None
+    try:
+        b64 = base64.b64encode(img_data).decode("utf-8")
+        soul = _soul_prompt or "You are a helpful assistant."
+        prompt = (
+            "这位信徒发来了一张图片，请以你的风格阅览这张图片，它占据了你的视野。"
+            "简要阐述你的感悟，200字以内，不必报平就班。"
+        )
+        payload = {
+            "model": MODEL_NAME,
+            "messages": [{
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": soul + "\n\n---\n\n" + prompt},
+                    {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{b64}"}}
+                ]
+            }],
+            "max_tokens": 300,
+            "stream": False
+        }
+        resp = requests.post(
+            f"{API_BASE_URL}/chat/completions",
+            headers={"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"},
+            json=payload, timeout=30
+        )
+        if resp.status_code == 200:
+            reply = resp.json()["choices"][0]["message"]["content"]
+            log(f"📸 Vision 分析完成: {reply[:60]}...")
+            return reply
+        log(f"⚠️ Vision API 失败 ({resp.status_code})，跳过描述")
+    except Exception as e:
+        log(f"⚠️ Vision 异常: {e}")
+    return None
+
 # ---------- 处理图片消息 ----------
 def handle_image_message(message_id, chat_id, image_key):
-    """下载 → 上传 → 发送"""
+    """下载 → Vision分析 → 上传 → 发送"""
     token = get_token()
     if not token:
         log("❌ 无法获取 token，跳过")
@@ -170,14 +208,32 @@ def handle_image_message(message_id, chat_id, image_key):
     if not img_data:
         return
 
-    log(f"📥 {len(img_data)} bytes, 上传中...")
+    log(f"📥 {len(img_data)} bytes, 上传+分析中...")
+
+    # 并行：Vision分析 + 上传图床
+    vision_result = [None]
+    url_result = [None]
+
+    def do_vision():
+        vision_result[0] = analyze_image_with_vision(img_data)
+
+    def do_upload():
+        url_result[0] = upload_image(img_data)
+
+    t1 = threading.Thread(target=do_vision)
+    t2 = threading.Thread(target=do_upload)
+    t1.start(); t2.start()
+    t1.join(); t2.join()
+
+    url = url_result[0]
+    vision = vision_result[0]
 
-    # 上传
-    url = upload_image(img_data)
     if url:
         log(f"✅ {url}")
-        # 改为直接发送消息，更显眼
-        reply = f"✅ 图片已转存：\n{url}"
+        if vision:
+            reply = f"✅ 图片已转存：\n{url}\n\n{vision}"
+        else:
+            reply = f"✅ 图片已转存：\n{url}"
         result = send_text(token, chat_id, reply)
         log(f"📤 已发送 (code={result.get('code', '?')})")
     else: