refactor: 重构项目配置管理，统一使用.env配置

- 新增config.py统一读取.env配置，移除硬编码路径和参数
- 重构collect_jrxml.py支持命令行参数和环境变量配置源目录
- 新增.env.example示例配置文件，整理所有可配置项
- 重构down_embedding_model.py、import_to_chroma.py等所有脚本使用统一配置
- 新增Windows一键部署脚本setup.bat
- 修正jrxml_banch_chunker.py的文件名拼写错误

embed_chunks.py

@@ -1,7 +1,7 @@
 """
 embed_chunks.py
-使用本地 Qwen3-Embedding-4B 模型对 JRXML chunks 进行向量化
-支持 GPU (CUDA) 或 CPU
+使用嵌入模型对 JRXML chunks 进行向量化
+支持 GPU (CUDA) 或 CPU，模型通过 .env / config.py 配置
 """
 
 import os
@@ -12,6 +12,10 @@ from pathlib import Path
 import numpy as np
 import torch
 from sentence_transformers import SentenceTransformer
+from config import (
+    EMBEDDING_MODEL_PATH, CHUNKER_OUTPUT_DIR, EMBEDDINGS_DIR,
+    USE_FP16, BATCH_SIZE, resolve_model_path
+)
 
 def build_text_for_embedding(chunk: dict) -> str:
     """
@@ -48,8 +52,8 @@ def build_text_for_embedding(chunk: dict) -> str:
 
 
 def main(chunks_json_path: str = None, output_dir: str = None,
-         model_path: str = None, batch_size: int = 64, normalize: bool = True,
-         use_fp16: bool = True):
+         model_path: str = None, batch_size: int = None, normalize: bool = True,
+         use_fp16: bool = None):
     """
     主流程：
     1. 加载 chunk JSON
@@ -60,19 +64,25 @@ def main(chunks_json_path: str = None, output_dir: str = None,
     project_root = Path(__file__).resolve().parent
 
     if chunks_json_path is None:
-        chunks_json_path = project_root / "jrxml_chunker_output" / "all_chunks.json"
+        chunks_json_path = CHUNKER_OUTPUT_DIR / "all_chunks.json"
     else:
         chunks_json_path = Path(chunks_json_path)
 
     if output_dir is None:
-        output_dir = project_root / "embeddings"
+        output_dir = EMBEDDINGS_DIR
     else:
         output_dir = Path(output_dir)
 
     if model_path is None:
-        model_path = project_root / "models" / "Qwen3-Embedding-4B"
+        model_path = resolve_model_path()
     else:
-        model_path = Path(model_path)
+        model_path = str(model_path)
+
+    if batch_size is None:
+        batch_size = BATCH_SIZE
+
+    if use_fp16 is None:
+        use_fp16 = USE_FP16
 
     if not chunks_json_path.exists():
         print(f"❌ Chunks 文件不存在: {chunks_json_path}")
@@ -91,19 +101,16 @@ def main(chunks_json_path: str = None, output_dir: str = None,
     print(f"\n🧠 加载嵌入模型: {model_path}")
     print(f"   设备: {device}")
 
-    # 检查是否是 HuggingFace Hub 模型（格式为 username/model_name）
     model_path_str = str(model_path)
-    # Windows PowerShell 会把 / 自动转成 \，需要还原
     if "\\" in model_path_str and not os.path.exists(model_path_str):
         model_path_str = model_path_str.replace("\\", "/")
-    
+
     is_hub_model = "/" in model_path_str and not os.path.exists(model_path_str)
-    
-    # 如果是本地路径但不存在，则报错
+
     if not is_hub_model and not os.path.exists(model_path_str):
         print(f"❌ 模型目录不存在: {model_path}")
-        print(f"   请先下载模型到 {model_path}")
-        print(f"   或者使用 HuggingFace Hub 模型，例如: sentence-transformers/all-MiniLM-L6-v2")
+        print(f"   请先运行 down_embedding_model.py 下载模型")
+        print(f"   或在 .env 中配置 EMBEDDING_MODEL_NAME 为 Hub 模型名")
         return None
 
     model = SentenceTransformer(model_path_str, device=device)
@@ -183,17 +190,17 @@ def main(chunks_json_path: str = None, output_dir: str = None,
 if __name__ == "__main__":
     import argparse
     project_root = Path(__file__).resolve().parent
-    default_chunks = project_root / "jrxml_chunker_output" / "all_chunks.json"
+    default_chunks = CHUNKER_OUTPUT_DIR / "all_chunks.json"
 
     parser = argparse.ArgumentParser(description="JRXML Chunks 向量化工具")
     parser.add_argument("chunks_json", nargs="?", default=str(default_chunks),
                         help=f"Chunks JSON 文件路径 (默认: {default_chunks})")
     parser.add_argument("--output_dir", "-o", default=None,
-                        help="输出目录 (默认: embeddings)")
+                        help=f"输出目录 (默认: {EMBEDDINGS_DIR})")
     parser.add_argument("--model_path", "-m", default=None,
-                        help="模型路径 (默认: models/Qwen3-Embedding-4B)")
-    parser.add_argument("--batch_size", "-b", type=int, default=64,
-                        help="批处理大小 (默认: 64)")
+                        help=f"模型路径 (默认: {resolve_model_path()})")
+    parser.add_argument("--batch_size", "-b", type=int, default=BATCH_SIZE,
+                        help=f"批处理大小 (默认: {BATCH_SIZE})")
     parser.add_argument("--no_normalize", action="store_true",
                         help="不做向量归一化")
     parser.add_argument("--no_fp16", action="store_true",