fix: 修复 11 个代码架构审计问题

H1: AppConfig 自定义 __init__ 改用 from_dict() 类方法
H2: list_collections_with_stats 改为从 ChromaDB 直接查询
H3: RateLimiter 过期 key 自动清理, 防止内存泄漏
M1: delete_by_source 区分 ValueError 与真实异常, 记日志
M2: VectorDB 新增 list_collections() 封装方法
M3: _remove_by_source 异常记日志, 不再静默吞掉
M4: CLI 集合回退支持 MD_VECTOR_DB_COLLECTION 环境变量
L1: DEFAULT_CONFIG_PATH 自动从项目根目录解析
L2: ingest 命令内重复 import 移至模块顶部
L3: 新增死循环回归测试 + 密集分隔符分块测试
L4: 统一 logger 名称为 md-vector-db
删除旧版审计文档

测试: 48 passed

src/cli/main.py

@@ -3,6 +3,8 @@ from pathlib import Path
 import sys
 import io
 import json
+import os
+import glob as _glob
 from typing import Annotated
 
 import typer
@@ -45,7 +47,10 @@ def _init_shared(config_path: str = DEFAULT_CONFIG_PATH):
 
 
 def _get_default_collection() -> str:
-    return _cfg.chroma.collection_name if _cfg else "markdown_docs"
+    return os.environ.get(
+        "MD_VECTOR_DB_COLLECTION",
+        _cfg.chroma.collection_name if _cfg else "markdown_docs",
+    )
 
 
 def _resolve_collection(collection: str | None) -> str:
@@ -94,10 +99,8 @@ def ingest(
         total = 0
         for fp in file_paths:
             # 支持通配符 (shell 展开或 Python glob)
-            from pathlib import Path as _Path
-            p = _Path(fp)
+            p = Path(fp)
             if "*" in fp or "?" in fp:
-                import glob as _glob
                 matches = _glob.glob(fp, recursive=True)
                 for m in matches:
                     c = ingestor.ingest_file(m)

src/core/config.py

@@ -72,19 +72,30 @@ class AppConfig:
     chunk: ChunkConfig = field(default_factory=ChunkConfig)
     server: ServerConfig = field(default_factory=ServerConfig)
 
-    def __init__(self, **kwargs):
-        self.chroma = ChromaConfig(**kwargs.get("chroma", {}))
-        self.embed = EmbedConfig(**kwargs.get("embed", {}))
-        self.chunk = ChunkConfig(**kwargs.get("chunk", {}))
-        self.server = ServerConfig(**kwargs.get("server", {}))
+    @classmethod
+    def from_dict(cls, data: dict) -> "AppConfig":
+        """从字典构造 (用于 YAML 加载)."""
+        return cls(
+            chroma=ChromaConfig(**data.get("chroma", {})),
+            embed=EmbedConfig(**data.get("embed", {})),
+            chunk=ChunkConfig(**data.get("chunk", {})),
+            server=ServerConfig(**data.get("server", {})),
+        )
 
 
 def load_config(path: str | None = None) -> AppConfig:
     """从 YAML 文件加载配置, 若文件不存在则返回默认配置."""
     config_path = path or DEFAULT_CONFIG_PATH
-    if not Path(config_path).exists():
+    config_file = Path(config_path)
+    # 相对路径 → 从项目根目录解析
+    if not config_file.is_absolute():
+        root = Path(__file__).resolve().parent.parent.parent
+        _try = root / config_file
+        if _try.exists():
+            config_file = _try
+    if not config_file.exists():
         return AppConfig()
 
-    with open(config_path, "r", encoding="utf-8") as f:
+    with open(config_file, "r", encoding="utf-8") as f:
         data = yaml.safe_load(f) or {}
-    return AppConfig(**data)
+    return AppConfig.from_dict(data)

src/core/db.py

@@ -20,6 +20,10 @@ class VectorDB:
         """获取或创建 collection."""
         return self.client.get_or_create_collection(name=name)
 
+    def list_collections(self) -> list[Collection]:
+        """列出所有 collection."""
+        return self.client.list_collections()
+
     def delete_collection(self, name: str) -> None:
         """删除 collection (线程安全)."""
         with self._write_lock:

src/core/ingest.py

@@ -1,4 +1,5 @@
 """Markdown 文档解析与入库模块."""
+import logging
 import re
 from dataclasses import dataclass
 from pathlib import Path
@@ -6,6 +7,8 @@ from pathlib import Path
 from src.core.db import VectorDB
 from src.core.embedder import Embedder, batch_embed
 
+logger = logging.getLogger("md-vector-db")
+
 
 class MarkdownSplitter:
     """Markdown 混合分块器：先按标题拆，超长再按段落拆."""
@@ -242,5 +245,7 @@ class DocumentIngestor:
                 )
                 if existing and existing["ids"]:
                     self.collection.delete(ids=existing["ids"])
+        except ValueError:
+            pass  # collection 为空时 ChromaDB 抛 ValueError
         except Exception:
-            pass  # collection 为空时 get 可能抛异常
+            logger.exception("去重检查失败: %s", file_name)

src/core/search.py

@@ -1,7 +1,10 @@
 """语义检索模块."""
+import logging
 from src.core.db import VectorDB
 from src.core.embedder import Embedder
 
+logger = logging.getLogger("md-vector-db")
+
 
 class Searcher:
     """向量检索器."""
@@ -84,6 +87,8 @@ class Searcher:
                 if existing and existing["ids"]:
                     self.collection.delete(ids=existing["ids"])
                     return True
+        except ValueError:
+            pass  # collection 为空时 ChromaDB 抛 ValueError
         except Exception:
-            pass
+            logger.exception("删除文档失败: %s", file_name)
         return False

src/server/auth.py

@@ -32,6 +32,8 @@ class RateLimiter:
         with self._lock:
             records = self._store[client_id]
             records[:] = [t for t in records if now - t < self.window]
+            if not records:
+                del self._store[client_id]  # 清理空 key，防止内存泄漏
             if len(records) >= self.max_requests:
                 return False
             records.append(now)

src/server/deps.py

@@ -8,7 +8,7 @@ from src.core.embedder import create_embedder
 from src.core.ingest import DocumentIngestor
 from src.core.search import Searcher
 
-logger = logging.getLogger(__name__)
+logger = logging.getLogger("md-vector-db")
 
 
 class AppState:
@@ -45,16 +45,13 @@ class AppState:
         return self._ingestors[name]
 
     def list_collections_with_stats(self) -> list[dict]:
-        """列出所有 collection 及其统计."""
+        """列出所有 collection 及其统计（直接从 ChromaDB 查询）."""
         result = []
-        all_names = set(self._searchers.keys()) | set(self._ingestors.keys())
-        all_names.add(self.default_collection)
-        for name in sorted(all_names):
-            try:
-                coll = self.db.get_or_create_collection(name)
-                result.append({"name": name, "count": coll.count()})
-            except Exception:
-                result.append({"name": name, "count": 0})
+        try:
+            for coll in self.db.client.list_collections():
+                result.append({"name": coll.name, "count": coll.count()})
+        except Exception:
+            logger.exception("列出集合失败")
         return result
 
     def is_healthy(self) -> dict: