feat: 添加Markdown分块器与统一批量分块入口，支持增量向量化与导入

- 新增 md_chunker.py: Markdown语义分块引擎，支持标题/代码块/表格智能拆分
- 新增 batch_chunker.py: 统一批量分块入口，支持JRXML+Markdown混合处理
- 新增 requirements.txt: 整理项目依赖
- embed_chunks.py: 新增 --incremental 增量模式，追加新向量到已有数据
- import_to_chroma.py: 新增 --incremental 增量模式，不再每次清空数据库
- 更新 README.md 与 docs/file_guide.md 反映最新架构

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

import_to_chroma.py

@@ -1,6 +1,7 @@
 """
 import_to_chroma.py
-将已生成的 chunk 向量导入 Chroma 数据库
+将 chunk 向量导入 Chroma 数据库
+支持 JRXML chunks 和 Markdown chunks 混合导入
 """
 
 import os
@@ -16,7 +17,8 @@ from config import EMBEDDINGS_DIR, CHROMA_DB_PATH, CHROMA_COLLECTION_NAME
 
 def main(embeddings_dir: str = None,
          chroma_path: str = None,
-         collection_name: str = None):
+         collection_name: str = None,
+         incremental: bool = False):
     """
     从 embeddings 目录读取向量和 chunks，导入 Chroma 持久化数据库
 
@@ -69,33 +71,55 @@ def main(embeddings_dir: str = None,
     chroma_path.mkdir(parents=True, exist_ok=True)
     client = chromadb.PersistentClient(path=str(chroma_path))
 
-    try:
-        client.delete_collection(collection_name)
-        print(f"   已删除旧集合 '{collection_name}'")
-    except Exception:
-        pass
-
-    collection = client.create_collection(
-        name=collection_name,
-        metadata={"hnsw:space": "cosine"}
-    )
+    if incremental:
+        try:
+            collection = client.get_collection(collection_name)
+            existing_ids = set(collection.get()['ids'])
+            print(f"   增量模式: 集合 '{collection_name}' 已有 {len(existing_ids)} 条记录")
+        except Exception:
+            collection = client.create_collection(
+                name=collection_name,
+                metadata={"hnsw:space": "cosine"}
+            )
+            existing_ids = set()
+            print(f"   增量模式: 创建新集合 '{collection_name}'")
+    else:
+        try:
+            client.delete_collection(collection_name)
+            print(f"   已删除旧集合 '{collection_name}'")
+        except Exception:
+            pass
+        collection = client.create_collection(
+            name=collection_name,
+            metadata={"hnsw:space": "cosine"}
+        )
+        existing_ids = set()
 
     print(f"\n🛠️ 准备导入数据...")
     ids = []
     documents = []
     metadatas = []
     embeddings_list = []
+    skipped = 0
 
     seen_ids = {}
     for i, chunk in enumerate(tqdm(chunks, desc="准备数据")):
         raw_id = str(chunk.get("chunk_id", i))
+        context = chunk.get("context", "")
+
         if raw_id in seen_ids:
             seen_ids[raw_id] += 1
-            chunk_id = f"{raw_id}_{seen_ids[raw_id]}"
+            unique_chunk_id = f"{raw_id}_{seen_ids[raw_id]}"
         else:
             seen_ids[raw_id] = 0
-            chunk_id = raw_id
-        ids.append(chunk_id)
+            unique_chunk_id = raw_id
+
+        # 增量模式：跳过已导入的
+        if incremental and unique_chunk_id in existing_ids:
+            skipped += 1
+            continue
+
+        ids.append(unique_chunk_id)
 
         doc_text = chunk.get("human_description", "")
         documents.append(doc_text)
@@ -105,7 +129,6 @@ def main(embeddings_dir: str = None,
         if chunk_type:
             meta["chunk_type"] = chunk_type
 
-        context = chunk.get("context", "")
         if context:
             meta["context"] = context
 
@@ -118,10 +141,26 @@ def main(embeddings_dir: str = None,
             meta["element_kind"] = chunk_meta["element_kind"]
         if "query_language" in chunk_meta:
             meta["query_language"] = chunk_meta["query_language"]
+        # Markdown-specific metadata
+        if "heading" in chunk_meta:
+            meta["heading"] = chunk_meta["heading"]
+        if "heading_level" in chunk_meta:
+            meta["heading_level"] = chunk_meta["heading_level"]
+        if "language" in chunk_meta:
+            meta["code_language"] = chunk_meta["language"]
 
         metadatas.append(meta)
         embeddings_list.append(embeddings[i].tolist())
 
+    if incremental and skipped > 0:
+        print(f"   增量模式: 跳过 {skipped} 条已存在记录")
+
+    if not ids:
+        print(f"\n✅ 没有新数据需要导入，集合已是最新")
+        print(f"   数据库路径: {chroma_path}")
+        print(f"   集合数量: {collection.count()}")
+        return collection
+
     print(f"\n📥 分批导入到 Chroma (每批 1000 条)...")
     import_batch_size = 1000
     start_time = time.time()
@@ -173,11 +212,14 @@ if __name__ == "__main__":
                         help=f"Chroma 数据库路径 (默认: {CHROMA_DB_PATH})")
     parser.add_argument("--collection_name", "-n", default=CHROMA_COLLECTION_NAME,
                         help=f"集合名称 (默认: {CHROMA_COLLECTION_NAME})")
+    parser.add_argument("--incremental", "-i", action="store_true",
+                        help="增量模式：只导入新增记录，不删除已有数据")
 
     args = parser.parse_args()
 
     main(
         embeddings_dir=args.embeddings_dir,
         chroma_path=args.chroma_path,
-        collection_name=args.collection_name
+        collection_name=args.collection_name,
+        incremental=args.incremental
     )