From b923e63bcde8148f92dbe6cc2f978f126f2ec174 Mon Sep 17 00:00:00 2001
From: lacerate551 <128470311+lacerate551@users.noreply.github.com>
Date: Tue, 9 Jun 2026 13:32:51 +0800
Subject: [PATCH] =?UTF-8?q?fix(knowledge):=20list=5Fcollections=20?=
 =?UTF-8?q?=E5=A2=9E=E5=8A=A0=E5=BC=82=E5=B8=B8=E4=BF=9D=E6=8A=A4=E4=B8=8E?=
 =?UTF-8?q?=E5=85=83=E6=95=B0=E6=8D=AE=E9=87=8D=E8=BD=BD?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

1. 遍历集合时 try/except 捕获 ChromaDB 异常（NotFoundError/InternalError），
   跳过失效集合并自动清理元数据，防止 /collections 500
2. 每次调用从磁盘重载 kb_metadata.json，确保多 worker 进程间状态一致，
   修复 GUNICORN_WORKERS>1 时返回数量波动问题
---
 knowledge/collection.py | 35 ++++++++++++++++++++++++++---------
 1 file changed, 26 insertions(+), 9 deletions(-)

diff --git a/knowledge/collection.py b/knowledge/collection.py
index ae06685..26b3ee1 100644
--- a/knowledge/collection.py
+++ b/knowledge/collection.py
@@ -348,6 +348,8 @@ class CollectionMixin:
             - department: 所属部门
             - description: 描述
         """
+        # 从磁盘重新加载元数据，确保多 worker 进程间状态一致
+        self._metadata = self._load_metadata()
         result = []
 
         # 扫描 base_path 下的所有子目录作为向量库
@@ -382,16 +384,31 @@ class CollectionMixin:
 
         self._save_metadata()
 
+        stale_collections = []
+
         for name, info in self._metadata.get("collections", {}).items():
-            collection = self.get_collection(name)
-            result.append(CollectionInfo(
-                name=name,
-                display_name=info.get("display_name", name),
-                document_count=collection.count() if collection else 0,
-                created_at=info.get("created_at", ""),
-                department=info.get("department", ""),
-                description=info.get("description", "")
-            ))
+            try:
+                collection = self.get_collection(name)
+                result.append(CollectionInfo(
+                    name=name,
+                    display_name=info.get("display_name", name),
+                    document_count=collection.count() if collection else 0,
+                    created_at=info.get("created_at", ""),
+                    department=info.get("department", ""),
+                    description=info.get("description", "")
+                ))
+            except Exception as e:
+                logger.warning(
+                    f"跳过异常向量库 '{name}': {e}，可能是 ChromaDB 数据目录已丢失"
+                )
+                stale_collections.append(name)
+
+        # 清理元数据中指向已失效集合的条目
+        if stale_collections:
+            for name in stale_collections:
+                self._metadata.get("collections", {}).pop(name, None)
+                logger.info(f"清理失效向量库元数据: {name}")
+            self._save_metadata()
 
         return result