feat: multi-collection support — each project uses its own isolated collection

src/server/deps.py

@@ -12,7 +12,7 @@ logger = logging.getLogger(__name__)
 
 
 class AppState:
-    """应用级共享状态."""
+    """应用级共享状态 — db 和 embedder 全局共享, searcher/ingestor 按 collection 懒创建."""
 
     def __init__(self):
         config_path = os.environ.get("MD_VECTOR_CONFIG", "config.yaml")
@@ -22,20 +22,48 @@ class AppState:
             "MD_VECTOR_DB_DATA_DIR", self.config.chroma.persist_dir
         )
         self.db = VectorDB(persist_dir=data_dir)
-
         self.embedder = create_embedder(self.config.embed)
 
-        collection = os.environ.get(
+        self.default_collection = os.environ.get(
             "MD_VECTOR_DB_COLLECTION", self.config.chroma.collection_name
         )
-        self.searcher = Searcher(self.db, self.embedder, collection)
-        self.ingestor = DocumentIngestor(self.db, self.embedder, collection)
+
+        # 按 collection 懒加载 searcher / ingestor
+        self._searchers: dict[str, Searcher] = {}
+        self._ingestors: dict[str, DocumentIngestor] = {}
+
+    def get_searcher(self, collection: str | None = None) -> Searcher:
+        name = collection or self.default_collection
+        if name not in self._searchers:
+            self._searchers[name] = Searcher(self.db, self.embedder, name)
+        return self._searchers[name]
+
+    def get_ingestor(self, collection: str | None = None) -> DocumentIngestor:
+        name = collection or self.default_collection
+        if name not in self._ingestors:
+            self._ingestors[name] = DocumentIngestor(self.db, self.embedder, name)
+        return self._ingestors[name]
+
+    def list_collections_with_stats(self) -> list[dict]:
+        """列出所有 collection 及其统计."""
+        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})
+        return result
 
     def is_healthy(self) -> dict:
         """真实健康检查: 验证 ChromaDB 和 Embedder 是否可用."""
         status = {"status": "ok", "checks": {}}
         try:
-            count = self.searcher.collection.count()
+            count = self.db.get_or_create_collection(
+                self.default_collection
+            ).count()
             status["checks"]["chromadb"] = {"status": "ok", "count": count}
         except Exception as e:
             status["checks"]["chromadb"] = {"status": "error", "detail": str(e)}
@@ -59,3 +87,8 @@ def get_state() -> AppState:
         logger.info("初始化应用状态...")
         _state = AppState()
     return _state
+
+
+def get_default_collection() -> str:
+    """获取默认 collection 名, 供 CLI 等非 HTTP 场景使用."""
+    return get_state().default_collection