diff --git a/mcp_docx.py b/mcp_docx.py
index 8b09e20..76d63bc 100644
--- a/mcp_docx.py
+++ b/mcp_docx.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 """
-docx_editor.py — 保留原格式替换文本 + 修改字体颜色 + 替换图片
+docx_editor.py — 保留原格式替换文本 + 修改字体颜色
 
 用法:
   # 列出文档中所有图片
@@ -10,25 +10,14 @@ docx_editor.py — 保留原格式替换文本 + 修改字体颜色 + 替换图
   python3 docx_editor.py input.docx output.docx \
     --replace "原文" "新文" \
     --color "关键词" "FF0000"
-
-  # 图片替换（按文档中出现的顺序，从1开始）
-  python3 docx_editor.py input.docx output.docx \
-    --image 1 new_chart.png \
-    --image 2 new_photo.jpg
-
-  # 同时替换文字和图片
-  python3 docx_editor.py input.docx output.docx \
-    --replace "旧标题" "新标题" \
-    --image 1 new_image.png \
-    --color "重点" "FF0000"
 """
 
 import argparse
+import copy
 import os
 import tempfile
 import zipfile
 from lxml import etree
-from PIL import Image
 import re
 
 W   = 'http://schemas.openxmlformats.org/wordprocessingml/2006/main'
@@ -37,12 +26,6 @@ A   = 'http://schemas.openxmlformats.org/drawingml/2006/main'
 R   = 'http://schemas.openxmlformats.org/officeDocument/2006/relationships'
 REL_TYPE_IMAGE = 'http://schemas.openxmlformats.org/officeDocument/2006/relationships/image'
 
-EXT_TO_MIME = {
-    'png': 'image/png', 'jpg': 'image/jpeg', 'jpeg': 'image/jpeg',
-    'gif': 'image/gif', 'bmp': 'image/bmp', 'tiff': 'image/tiff',
-    'webp': 'image/webp',
-}
-
 
 def unpack(docx_path, out_dir):
     """使用 zipfile 直接解包 .docx 到临时目录，替代外部 unpack.py 脚本。"""
@@ -141,61 +124,146 @@ def get_images_info(docx_path):
         return build_image_index(tmpdir)
 
 
-def replace_image(unpacked_dir, index, new_image_path):
-    """替换第 index 张图片（1-based）"""
-    imgs = build_image_index(unpacked_dir)
-    if index < 1 or index > len(imgs):
-        raise ValueError(f"图片序号 {index} 超出范围（共 {len(imgs)} 张）")
+def _normalize_newlines(text):
+    if text is None:
+        return ''
+    return str(text).replace('\r\n', '\n').replace('\r', '\n')
 
-    info     = imgs[index - 1]
-    old_abs  = info['abs_path']
-    old_ext  = info['ext']
-    new_ext  = os.path.splitext(new_image_path)[1].lstrip('.').lower()
-    if new_ext == 'jpg':
-        new_ext = 'jpeg'
 
-    print(f"    图片#{index} {os.path.basename(info['media_file'])}({old_ext.upper()})"
-          f" ← {os.path.basename(new_image_path)}({new_ext.upper()})")
+def _is_text_node(el):
+    return el.tag == f'{{{W}}}t'
 
-    if old_ext == new_ext:
-        # ── 同格式：直接覆盖 ──────────────────────────────
-        import shutil
-        shutil.copy2(new_image_path, old_abs)
 
+def _is_break_node(el):
+    return el.tag in (f'{{{W}}}br', f'{{{W}}}cr')
+
+
+def _is_tab_node(el):
+    return el.tag == f'{{{W}}}tab'
+
+
+def _iter_run_text_parts(run_el):
+    for child in run_el:
+        if _is_text_node(child):
+            yield child, _normalize_newlines(child.text or '')
+        elif _is_break_node(child):
+            yield child, '\n'
+        elif _is_tab_node(child):
+            yield child, '\t'
+
+
+def _run_text(run_el):
+    return ''.join(part for _, part in _iter_run_text_parts(run_el))
+
+
+def _paragraph_text(para_el):
+    return ''.join(_run_text(run) for run in para_el.iter(f'{{{W}}}r'))
+
+
+def _clear_run_text_like_children(run_el):
+    for child in list(run_el):
+        if _is_text_node(child) or _is_break_node(child) or _is_tab_node(child):
+            run_el.remove(child)
+
+
+def _append_text_to_run(run_el, text):
+    text = _normalize_newlines(text)
+    parts = text.split('\n')
+
+    if len(parts) == 1:
+        t_el = etree.SubElement(run_el, f'{{{W}}}t')
+        t_el.text = parts[0]
+        if parts[0] and (parts[0][0] == ' ' or parts[0][-1] == ' '):
+            t_el.set('{http://www.w3.org/XML/1998/namespace}space', 'preserve')
+        return
+
+    for idx, part in enumerate(parts):
+        if part:
+            t_el = etree.SubElement(run_el, f'{{{W}}}t')
+            t_el.text = part
+            if part[0] == ' ' or part[-1] == ' ':
+                t_el.set('{http://www.w3.org/XML/1998/namespace}space', 'preserve')
+        if idx < len(parts) - 1:
+            etree.SubElement(run_el, f'{{{W}}}br')
+
+
+def _ensure_paragraph_run(para_el):
+    runs = list(para_el.findall(f'.//{{{W}}}r'))
+    if runs:
+        return runs[0]
+
+    ppr = para_el.find(f'{{{W}}}pPr')
+    new_r = etree.Element(f'{{{W}}}r')
+    if ppr is None:
+        para_el.insert(0, new_r)
     else:
-        # ── 不同格式：Pillow 转换 + 更新 rels + ContentTypes
-        new_abs = os.path.splitext(old_abs)[0] + '.' + new_ext
-        img = Image.open(new_image_path)
-        fmt = {'jpeg': 'JPEG', 'png': 'PNG', 'gif': 'GIF',
-               'bmp': 'BMP', 'tiff': 'TIFF', 'webp': 'WEBP'}.get(new_ext, new_ext.upper())
-        if fmt == 'JPEG' and img.mode in ('RGBA', 'P'):
-            img = img.convert('RGB')
-        img.save(new_abs, format=fmt)
-        if os.path.abspath(new_abs) != os.path.abspath(old_abs):
-            os.remove(old_abs)
+        para_el.insert(para_el.index(ppr) + 1, new_r)
+    return new_r
 
-        # 更新 rels
-        old_media = info['media_file']
-        new_media = os.path.splitext(old_media)[0] + '.' + new_ext
-        word_dir  = os.path.join(unpacked_dir, 'word')
-        rels_path = os.path.join(word_dir, '_rels', 'document.xml.rels')
-        rels_tree = etree.parse(rels_path)
-        for rel in rels_tree.getroot():
-            if rel.get('Id') == info['rid']:
-                rel.set('Target', new_media)
-                break
-        rels_tree.write(rels_path, xml_declaration=True, encoding='UTF-8', standalone=True)
 
-        # 更新 ContentTypes
-        ct_path = os.path.join(unpacked_dir, '[Content_Types].xml')
-        ct_tree = etree.parse(ct_path)
-        ct_root = ct_tree.getroot()
-        existing = {el.get('Extension', '') for el in ct_root}
-        if new_ext not in existing:
-            etree.SubElement(ct_root, 'Default', Extension=new_ext,
-                             ContentType=EXT_TO_MIME.get(new_ext, f'image/{new_ext}'))
-        ct_tree.write(ct_path, xml_declaration=True, encoding='UTF-8', standalone=True)
-        print(f"      格式转换 {old_ext}→{new_ext}，rels 和 ContentTypes 已更新")
+def _set_paragraph_text(para_el, text):
+    runs = list(para_el.findall(f'.//{{{W}}}r'))
+    text_runs = [run for run in runs if any(True for _ in _iter_run_text_parts(run))]
+
+    if text_runs:
+        first_run = text_runs[0]
+        for run in text_runs:
+            _clear_run_text_like_children(run)
+    else:
+        first_run = _ensure_paragraph_run(para_el)
+        _clear_run_text_like_children(first_run)
+
+    _append_text_to_run(first_run, text)
+
+
+def _paragraph_list(doc_el):
+    return list(doc_el.iter(f'{{{W}}}p'))
+
+
+def _replace_paragraph_block(doc_el, old_text, new_text):
+    old_segments = _normalize_newlines(old_text).split('\n\n')
+    new_segments = _normalize_newlines(new_text).split('\n\n')
+    if len(old_segments) <= 1:
+        return False
+
+    paras = _paragraph_list(doc_el)
+    para_texts = [_paragraph_text(p) for p in paras]
+
+    match_start = None
+    for i in range(0, len(para_texts) - len(old_segments) + 1):
+        if para_texts[i:i + len(old_segments)] == old_segments:
+            match_start = i
+            break
+
+    if match_start is None:
+        return False
+
+    matched_paras = paras[match_start:match_start + len(old_segments)]
+    parent = matched_paras[0].getparent()
+    if parent is None:
+        return False
+
+    anchor_index = parent.index(matched_paras[-1])
+
+    shared_count = min(len(matched_paras), len(new_segments))
+    for idx in range(shared_count):
+        _set_paragraph_text(matched_paras[idx], new_segments[idx])
+
+    if len(new_segments) > len(matched_paras):
+        template_para = matched_paras[-1]
+        insert_at = anchor_index + 1
+        for seg in new_segments[len(matched_paras):]:
+            new_para = copy.deepcopy(template_para)
+            _set_paragraph_text(new_para, seg)
+            parent.insert(insert_at, new_para)
+            insert_at += 1
+    elif len(new_segments) < len(matched_paras):
+        for para in matched_paras[len(new_segments):]:
+            para_parent = para.getparent()
+            if para_parent is not None:
+                para_parent.remove(para)
+
+    return True
 
 
 def paragraph_replace(para_el, replacements):
@@ -213,20 +281,27 @@ def paragraph_replace(para_el, replacements):
         return
 
     # 收集所有文本元素及其位置信息
-    t_elements = []
+    text_runs = []
     for run in runs:
-        for t_el in run.findall(f'{{{W}}}t'):
-            t_elements.append((run, t_el))
+        if any(True for _ in _iter_run_text_parts(run)):
+            text_runs.append(run)
 
-    if not t_elements:
+    if not text_runs:
         return
 
     # 拼接完整文本
-    full_text = ''.join(t_el.text or '' for _, t_el in t_elements)
+    full_text = _paragraph_text(para_el)
     original_text = full_text
 
-    # 执行所有替换
+    normalized_replacements = []
     for old, new in replacements:
+        normalized_replacements.append((
+            _normalize_newlines(old),
+            _normalize_newlines(new),
+        ))
+
+    # 执行所有替换
+    for old, new in normalized_replacements:
         if old in full_text:
             full_text = full_text.replace(old, new)
 
@@ -236,16 +311,11 @@ def paragraph_replace(para_el, replacements):
 
     print(f"段落替换: {len(original_text)} -> {len(full_text)} 字符")
 
-    # 将新文本重新分配到原有的 <w:t> 元素中
-    # 策略：将所有文本放入第一个元素，清空其他元素，避免不当切分导致换行
-    _, first_t_el = t_elements[0]
-    first_t_el.text = full_text
-    if full_text and (full_text[0] == ' ' or full_text[-1] == ' '):
-        first_t_el.set('{http://www.w3.org/XML/1998/namespace}space', 'preserve')
-
-    # 清空其他 <w:t> 元素
-    for i in range(1, len(t_elements)):
-        t_elements[i][1].text = ''
+    # 将规范化文本重新写回第一个文本 run，\n 会回写成 Word 的换行节点。
+    first_run = text_runs[0]
+    for run in text_runs:
+        _clear_run_text_like_children(run)
+    _append_text_to_run(first_run, full_text)
 
 
 def ensure_rpr(run_el):
@@ -271,13 +341,15 @@ def apply_color_to_keyword(doc_el, keyword, hex_color, context_text=None):
     当 context_text 不为空时，只在“整段文本包含该 context_text 的段落”中进行上色，
     避免同一个关键字在其他段落里被误伤（例如单独的数字 0）。
     """
+    keyword = _normalize_newlines(keyword)
+    context_text = _normalize_newlines(context_text) if context_text is not None else None
+
     # 如果提供了上下文，只在包含该上下文的段落内着色
     allowed_paras = None
     if context_text:
         allowed_paras = set()
         for p in doc_el.iter(f'{{{W}}}p'):
-            t_nodes = list(p.iter(f'{{{W}}}t'))
-            full = ''.join(t.text or '' for t in t_nodes)
+            full = _paragraph_text(p)
             if context_text in full:
                 allowed_paras.add(p)
 
@@ -294,10 +366,9 @@ def apply_color_to_keyword(doc_el, keyword, hex_color, context_text=None):
             para = _find_ancestor_para(run)
             if para not in allowed_paras:
                 continue
-        t_nodes = list(run.findall(f'{{{W}}}t'))
-        if not t_nodes:
+        full_text = _run_text(run)
+        if not full_text:
             continue
-        full_text = ''.join(t.text or '' for t in t_nodes)
         if keyword not in full_text:
             continue
 
@@ -317,10 +388,7 @@ def apply_color_to_keyword(doc_el, keyword, hex_color, context_text=None):
             new_r = etree.Element(f'{{{W}}}r')
             if rpr_bytes is not None:
                 new_r.append(etree.fromstring(rpr_bytes))
-            t_el = etree.SubElement(new_r, f'{{{W}}}t')
-            t_el.text = text
-            if text and (text[0] == ' ' or text[-1] == ' '):
-                t_el.set('{http://www.w3.org/XML/1998/namespace}space', 'preserve')
+            _append_text_to_run(new_r, text)
             if colored:
                 set_color_on_rpr(ensure_rpr(new_r), hex_color)
             return new_r
@@ -405,19 +473,13 @@ def remove_rule_blocks(doc_el):
         if parent is not None:
             parent.remove(p)
 
-def process(input_docx, output_docx, replacements, image_replacements,
-            color_keywords):
+def process(input_docx, output_docx, replacements, color_keywords):
     with tempfile.TemporaryDirectory() as tmpdir:
         print(f"📂 解包 {input_docx} ...")
         unpack(input_docx, tmpdir)
 
         doc_xml_path = os.path.join(tmpdir, 'word', 'document.xml')
 
-        if image_replacements:
-            print(f"🖼️  替换 {len(image_replacements)} 张图片...")
-            for idx, new_img in image_replacements:
-                replace_image(tmpdir, idx, new_img)
-
         tree = etree.parse(doc_xml_path)
         root = tree.getroot()
 
@@ -426,8 +488,17 @@ def process(input_docx, output_docx, replacements, image_replacements,
 
         if replacements:
             print(f"✏️  替换 {len(replacements)} 条文本...")
-            for para in root.iter(f'{{{W}}}p'):
-                paragraph_replace(para, replacements)
+            remaining_replacements = []
+            for old, new in replacements:
+                if '\n\n' in _normalize_newlines(old):
+                    replaced = _replace_paragraph_block(root, old, new)
+                    if replaced:
+                        print("🧩 跨段替换命中")
+                        continue
+                remaining_replacements.append((old, new))
+            if remaining_replacements:
+                for para in root.iter(f'{{{W}}}p'):
+                    paragraph_replace(para, remaining_replacements)
 
         # 根据 span 解析出的关键字上色
         for item in color_keywords:
@@ -457,6 +528,8 @@ def _parse_span_replacement(new_text):
     """
     import re
 
+    new_text = _normalize_newlines(new_text)
+
     # 简单的命名颜色到 16 进制的映射，可按需扩展
     named_colors = {
         'red': 'FF0000',
@@ -505,33 +578,33 @@ def _parse_span_replacement(new_text):
         re.IGNORECASE | re.DOTALL,
     )
 
-    # 先得到去掉 span 标签后的纯文本（也是最终会写入 DOCX 的内容）
+    # 先按段落边界拆分，这样 span 上色时可以使用所在段落作为上下文。
     def _strip_repl(m):
         return m.group(2)
 
-    plain_text = span_pattern.sub(_strip_repl, new_text)
-
-    # 再次遍历 span，收集颜色关键字，并把“整句纯文本”作为上下文挂在每个关键字上
+    plain_segments = []
     color_keywords = []
-    for m in span_pattern.finditer(new_text):
-        raw_color = m.group(1)
-        hex_color = _normalize_color(raw_color)
-        keyword = m.group(2)
-        # 三元组: (关键字, 颜色, 该 NEW 对应的整句纯文本上下文)
-        color_keywords.append((keyword, hex_color, plain_text))
+    for segment in new_text.split('\n\n'):
+        plain_segment = span_pattern.sub(_strip_repl, segment)
+        plain_segments.append(plain_segment)
+        for m in span_pattern.finditer(segment):
+            raw_color = m.group(1)
+            hex_color = _normalize_color(raw_color)
+            keyword = m.group(2)
+            # 三元组: (关键字, 颜色, 所在段落的纯文本上下文)
+            color_keywords.append((keyword, hex_color, plain_segment))
 
+    plain_text = '\n\n'.join(plain_segments)
     return plain_text, color_keywords
 
 
 def main():
-    parser = argparse.ArgumentParser(description='DOCX 格式保留：替换文本/图片/颜色')
+    parser = argparse.ArgumentParser(description='DOCX 格式保留：替换文本/颜色')
     parser.add_argument('input', help='输入 .docx')
     parser.add_argument('output', nargs='?', help='输出 .docx')
     parser.add_argument('--list-images', action='store_true', help='列出所有图片')
     parser.add_argument('--replace', nargs=2, metavar=('OLD', 'NEW'),
                         action='append', default=[])
-    parser.add_argument('--image', nargs=2, metavar=('INDEX', 'FILE'),
-                        action='append', default=[], help='图片替换')
     args = parser.parse_args()
 
     if args.list_images:
@@ -549,12 +622,11 @@ def main():
         color_keywords.extend(spans)
 
     process(
-        input_docx        = args.input,
-        output_docx       = args.output,
-        replacements      = replacements,
-        image_replacements= [(int(i), f) for i, f in args.image],
-        color_keywords    = color_keywords,
+        input_docx=args.input,
+        output_docx=args.output,
+        replacements=replacements,
+        color_keywords=color_keywords,
     )
 
 if __name__ == '__main__':
-    main()
\ No newline at end of file
+    main()
diff --git a/mcp_docx_server.py b/mcp_docx_server.py
index 364c4f4..b3986c6 100644
--- a/mcp_docx_server.py
+++ b/mcp_docx_server.py
@@ -2,9 +2,9 @@
 """
 基于 mcp_docx.py 封装的 MCP 服务器。
 
-暴露两个主要 MCP 工具：
+暴露主要 MCP 工具：
 - list_docx_images：列出 DOCX 中的图片信息
-- edit_docx:       进行文本替换 / 关键字上色 / 图片替换
+- edit_docx：       进行文本替换 / 关键字上色（与 HTTP POST /edit_docx 能力一致）
 
 额外提供 HTTP 文件接口（仅在 http 模式下可用）：
 - POST /upload:    上传文件到服务器
@@ -29,12 +29,17 @@
 """
 
 import argparse
+import hashlib
+import json
 import os
+import shutil
 import tempfile
+import time
 import urllib.parse
+import zipfile
+from contextlib import contextmanager
 from datetime import datetime, date, timedelta
 from typing import Any, Dict, List, Optional
-import uuid
 
 import requests
 from lxml import etree
@@ -43,6 +48,7 @@ from mcp.server.transport_security import TransportSecuritySettings
 
 from mcp_docx import (
     W,
+    _normalize_newlines,
     get_images_info,
     process,
     _parse_span_replacement,
@@ -50,8 +56,18 @@ from mcp_docx import (
     unpack,
     pack,
 )
+# HTTP 远程模式：添加文件上传下载路由
+from starlette.responses import FileResponse, JSONResponse
+from starlette.background import BackgroundTask
+from starlette.requests import Request
+from starlette.responses import FileResponse, JSONResponse
 
-_disable_dns_rebinding = os.getenv("MCP_DISABLE_HOST_CHECK") == "1"
+if os.name == "nt":
+    import msvcrt
+else:
+    import fcntl
+
+_disable_dns_rebinding = True
 
 if _disable_dns_rebinding:
     # 参考 python-sdk 官方文档：关闭 DNS rebinding 防护（适合本地或已由外层网关做安全控制的环境）
@@ -64,8 +80,8 @@ else:
     # 如需通过网关 / 域名访问，可在这里追加 allowed_hosts / allowed_origins
     transport_security = TransportSecuritySettings(
         enable_dns_rebinding_protection=True,
-        allowed_hosts=["localhost:*", "127.0.0.1:*", "192.168.10.101:*"],
-        allowed_origins=["http://localhost:*", "http://127.0.0.1:*","http://192.168.10.101:*"],
+        allowed_hosts=["localhost:*", "127.0.0.1:*", "192.168.1.13:*","10.150.172.13:*"],
+        allowed_origins=["http://localhost:*", "http://127.0.0.1:*","http://192.168.1.13:*","http://10.150.172.13:*"],
     )
 
 
@@ -234,6 +250,53 @@ def _download_to_temp(url: str, suffix: str = ".tmp") -> str:
     return tmp_path
 
 
+def _safe_filename(filename: Optional[str], default: str = "uploaded.docx") -> str:
+    """提取安全文件名，避免路径穿越。"""
+    if not filename:
+        return default
+    decoded = urllib.parse.unquote(str(filename))
+    safe_name = os.path.basename(decoded).strip()
+    return safe_name or default
+
+
+def _filename_from_url(url: str, default: str = "uploaded.docx") -> str:
+    """从 URL 中推断文件名，优先读取 query 参数中的 filename。"""
+    parsed = urllib.parse.urlparse(url)
+    query = urllib.parse.parse_qs(parsed.query)
+
+    for key in ("filename", "fileName", "name"):
+        values = query.get(key)
+        if values:
+            return _safe_filename(values[0], default=default)
+
+    return _safe_filename(os.path.basename(parsed.path), default=default)
+
+
+def _download_to_path(url: str, local_path: str) -> None:
+    """将远程 URL 下载到指定路径，完成后原子覆盖目标文件。"""
+    resp = requests.get(url, stream=True, timeout=30)
+    resp.raise_for_status()
+
+    parent_dir = os.path.dirname(local_path) or "."
+    os.makedirs(parent_dir, exist_ok=True)
+    fd, tmp_path = tempfile.mkstemp(
+        suffix=os.path.splitext(local_path)[1] or ".tmp",
+        dir=parent_dir,
+    )
+    try:
+        with os.fdopen(fd, "wb") as f:
+            for chunk in resp.iter_content(chunk_size=8192):
+                if chunk:
+                    f.write(chunk)
+        os.replace(tmp_path, local_path)
+    except Exception:
+        try:
+            os.remove(tmp_path)
+        except OSError:
+            pass
+        raise
+
+
 def _build_output_url(abs_output_path: str) -> Optional[str]:
     """
     构造输出文件的下载 URL。
@@ -245,18 +308,37 @@ def _build_output_url(abs_output_path: str) -> Optional[str]:
       - 否则在 http 模式下: http://host:port/download/{filename}
       - stdio 模式下: 返回 None
     """
+    filename = os.path.basename(abs_output_path)
+    encoded_filename = urllib.parse.quote(filename)
+
+    def _append_filename(base_url: str) -> str:
+        parsed = urllib.parse.urlparse(base_url)
+        query = urllib.parse.parse_qsl(parsed.query, keep_blank_values=True)
+
+        for index, (key, _) in enumerate(query):
+            if key in ("filename", "fileName", "name"):
+                query[index] = (key, filename)
+                return urllib.parse.urlunparse(
+                    parsed._replace(query=urllib.parse.urlencode(query))
+                )
+
+        if parsed.path.rstrip("/").endswith("/download"):
+            query.append(("filename", filename))
+            return urllib.parse.urlunparse(
+                parsed._replace(query=urllib.parse.urlencode(query))
+            )
+
+        return base_url.rstrip("/") + "/" + encoded_filename
+
     # 优先使用环境变量
     base = os.getenv("MCP_OUTPUT_BASE_URL")
     if base:
-        filename = os.path.basename(abs_output_path)
-        return base.rstrip("/") + "/" + filename
+        return _append_filename(base)
 
     # 如果是 http 模式，自动构建下载 URL
     if _server_config["transport"] == "http":
         host = _server_config["host"]
         port = _server_config["port"]
-        filename = os.path.basename(abs_output_path)
-
         # 如果 host 是 0.0.0.0，尝试使用更具体的地址
         if host == "0.0.0.0":
             # 优先使用环境变量指定的公网地址
@@ -267,7 +349,7 @@ def _build_output_url(abs_output_path: str) -> Optional[str]:
                 # 默认使用 localhost
                 host = "localhost"
 
-        return f"http://{host}:{port}/download/{filename}"
+        return _append_filename(f"http://{host}:{port}/download")
 
     return None
 
@@ -294,6 +376,136 @@ def _get_tmp_upload_dir() -> str:
     return os.path.abspath(tmp_dir)
 
 
+def _get_lock_dir() -> str:
+    """获取文件锁目录。"""
+    lock_dir = os.path.join(_get_upload_dir(), ".locks")
+    os.makedirs(lock_dir, exist_ok=True)
+    return lock_dir
+
+
+def _get_lock_path(target_path: str) -> str:
+    """根据目标文件路径生成稳定的锁文件路径。"""
+    abs_target = os.path.abspath(target_path)
+    base_name = _safe_filename(os.path.basename(abs_target), default="file")
+    digest = hashlib.sha256(abs_target.encode("utf-8")).hexdigest()
+    return os.path.join(_get_lock_dir(), f"{base_name}.{digest}.lock")
+
+
+def _acquire_lock(handle) -> None:
+    """跨进程独占锁。"""
+    handle.seek(0, os.SEEK_END)
+    if handle.tell() == 0:
+        handle.write(b"0")
+        handle.flush()
+    handle.seek(0)
+
+    if os.name == "nt":
+        while True:
+            try:
+                msvcrt.locking(handle.fileno(), msvcrt.LK_LOCK, 1)
+                break
+            except OSError:
+                time.sleep(0.05)
+    else:
+        fcntl.flock(handle.fileno(), fcntl.LOCK_EX)
+
+
+def _release_lock(handle) -> None:
+    """释放跨进程独占锁。"""
+    handle.seek(0)
+    if os.name == "nt":
+        msvcrt.locking(handle.fileno(), msvcrt.LK_UNLCK, 1)
+    else:
+        fcntl.flock(handle.fileno(), fcntl.LOCK_UN)
+
+
+@contextmanager
+def _file_lock(target_path: str):
+    """针对目标文件路径获取独占文件锁。"""
+    lock_path = _get_lock_path(target_path)
+    with open(lock_path, "a+b") as handle:
+        _acquire_lock(handle)
+        try:
+            yield
+        finally:
+            _release_lock(handle)
+
+
+def _write_bytes_atomic(file_path: str, content: bytes) -> None:
+    """原子写入文件内容。"""
+    parent_dir = os.path.dirname(file_path) or "."
+    os.makedirs(parent_dir, exist_ok=True)
+
+    fd, tmp_path = tempfile.mkstemp(
+        suffix=os.path.splitext(file_path)[1] or ".tmp",
+        dir=parent_dir,
+    )
+    try:
+        with os.fdopen(fd, "wb") as f:
+            f.write(content)
+        os.replace(tmp_path, file_path)
+    except Exception:
+        try:
+            os.remove(tmp_path)
+        except OSError:
+            pass
+        raise
+
+
+def _snapshot_file(file_path: str) -> str:
+    """复制文件快照，供下载接口在释放锁后返回。"""
+    suffix = os.path.splitext(file_path)[1] or ".tmp"
+    fd, snapshot_path = tempfile.mkstemp(suffix=suffix)
+    try:
+        with os.fdopen(fd, "wb") as dst, open(file_path, "rb") as src:
+            shutil.copyfileobj(src, dst)
+    except Exception:
+        try:
+            os.remove(snapshot_path)
+        except OSError:
+            pass
+        raise
+    return snapshot_path
+
+
+def _cleanup_temp_file(file_path: str) -> None:
+    try:
+        os.remove(file_path)
+    except OSError:
+        pass
+
+
+def _validate_docx_file(file_path: str) -> None:
+    if not os.path.exists(file_path):
+        raise FileNotFoundError(f"输入 DOCX 文件不存在: {file_path}")
+    if not os.path.isfile(file_path):
+        raise FileNotFoundError(f"输入路径不是文件: {file_path}")
+
+    size = os.path.getsize(file_path)
+    if size <= 0:
+        raise ValueError(f"输入 DOCX 文件为空: {os.path.basename(file_path)}")
+    if not zipfile.is_zipfile(file_path):
+        raise ValueError(f"输入文件不是合法的 DOCX/ZIP: {os.path.basename(file_path)}")
+
+
+def _resolve_edit_target_path(input_docx_path: str, upload_dir: str) -> str:
+    """
+    将编辑输入统一解析为本地路径。
+
+    - URL 输入会先下载到 uploads 目录
+    - 相对路径按 uploads 下的文件名处理
+    - 绝对路径直接使用
+    """
+    if _is_url(input_docx_path):
+        filename = _filename_from_url(input_docx_path)
+        return os.path.join(upload_dir, filename)
+
+    if os.path.isabs(input_docx_path):
+        return input_docx_path
+
+    return os.path.join(upload_dir, _safe_filename(input_docx_path))
+
+
 @mcp.tool()
 async def list_docx_images(docx_url: str) -> List[Dict[str, Any]]:
     """
@@ -313,72 +525,34 @@ async def list_docx_images(docx_url: str) -> List[Dict[str, Any]]:
     imgs = get_images_info(_download_to_temp(docx_url, suffix=".docx"))
     return imgs
 
-@mcp.custom_route("/edit_docx", methods=["POST"])
-async def edit_docx_handler(request: Request):
-    data = await request.json()
-    input_docx_path = data.get("input_docx_path")
-    replacements = data.get("replacements")
-    image_replacements = data.get("image_replacements")
-    report_type = data.get("report_type")
-    report_title_time = data.get("report_title_time")
+
+def _edit_docx_core(
+    input_docx_path: str,
+    replacements: Optional[List[Dict[str, Any]]],
+    report_type: Optional[str],
+    report_title_time: Optional[str],
+) -> Dict[str, Any]:
     """
-    对 DOCX 文件进行编辑。
-
-    支持：
-      - 纯文本替换
-      - 通过 <span color="red">关键字</span> 语法设置关键字颜色
-      - 替换指定序号的图片
-      - 报告日期与期数自动替换（仅在“目录”之前生效）
-
-    参数:
-      - input_docx_path: 输入 DOCX 文件名称
-      - replacements: 文本替换规则列表，例如:
-            [
-              {"old": "计划作业总数共有10项。", "new": "计划作业总数共有<span color='red'>XX</span>项。"},
-              {"old": "文档原文本，必须是完整的一句话或者段落", "new": "要替换的文本"}
-            ]
-      - image_replacements: 图片替换规则
-      - report_type: 报告类型，可选值：日报 / 周报 / 月报（或对应的英文 daily / weekly / monthly）
-      - report_title_time: 报告标题中要显示的时间字符串，用来替换“YYYY年M月”这一段（仅在第一次匹配时生效）
+    对 DOCX 文件进行编辑（与 HTTP /edit_docx 共用逻辑）。
 
     返回:
-      - {
-          "output_path": 生成的 DOCX 绝对路径,
-          "output_url":  如果配置了 MCP_OUTPUT_BASE_URL，则为可访问该文件的 URL，否则为 null
-        }
+      - {"output_path": 绝对路径, "output_url": URL 或 None}
     """
-    tmp_input: Optional[str] = None
-    tmp_images: List[str] = []
-    print(f"edit_docx: input_docx_path: {input_docx_path}, replacements: {replacements}, image_replacements: {image_replacements}")
+    print(f"edit_docx: input_docx_path: {input_docx_path}, replacements: {replacements}")
+    upload_dir = _get_upload_dir()
+    local_input = _resolve_edit_target_path(input_docx_path, upload_dir)
+    lock_cm = _file_lock(local_input)
+    lock_cm.__enter__()
     try:
-        upload_dir = _get_upload_dir()      # 输出目录：/uploads
-        tmp_upload_dir = _get_tmp_upload_dir()  # 上传临时目录：/tmp
 
-        # 解析输入路径：支持 URL、绝对路径、仅文件名三种形式
-        local_input = input_docx_path
         if _is_url(input_docx_path):
-            parsed = urllib.parse.urlparse(input_docx_path)
-            ext = os.path.splitext(parsed.path)[1] or ".docx"
-            tmp_input = _download_to_temp(input_docx_path, suffix=ext)
-            local_input = tmp_input
-        elif not os.path.isabs(local_input):
-            # 相对路径：优先在 tmp，其次在 uploads 中查找
-            cand_tmp = os.path.join(tmp_upload_dir, input_docx_path)
-            cand_upload = os.path.join(upload_dir, input_docx_path)
-            if os.path.exists(cand_tmp):
-                local_input = cand_tmp
-            else:
-                local_input = cand_upload
+            _download_to_path(input_docx_path, local_input)
 
-        if not os.path.exists(local_input):
-            raise FileNotFoundError(f"输入 DOCX 文件不存在: {input_docx_path}")
+        _validate_docx_file(local_input)
 
         if replacements is None:
             replacements = []
-        if image_replacements is None:
-            image_replacements = []
 
-        # 解析文本替换与颜色关键字（复用 CLI 逻辑）
         rep_pairs = []
         color_keywords = []
         for item in replacements:
@@ -386,55 +560,25 @@ async def edit_docx_handler(request: Request):
             new_raw = item.get("new")
             if not old:
                 continue
+            old = _normalize_newlines(old)
             if new_raw is None:
                 new_raw = ""
+            else:
+                new_raw = _normalize_newlines(new_raw)
             new_plain, spans = _parse_span_replacement(new_raw)
             rep_pairs.append((old, new_plain))
             color_keywords.extend(spans)
 
-        # 处理图片替换参数（支持本地路径或 URL）
-        img_pairs = []
-        for item in image_replacements:
-            try:
-                idx = int(item.get("index"))
-            except (TypeError, ValueError):
-                continue
-
-            path = item.get("file")
-            if not path:
-                continue
-
-            local_img = path
-            if _is_url(path):
-                parsed = urllib.parse.urlparse(path)
-                ext = os.path.splitext(parsed.path)[1] or ""
-                suffix = ext if ext else ".img"
-                tmp_img = _download_to_temp(path, suffix=suffix)
-                tmp_images.append(tmp_img)
-                local_img = tmp_img
-
-            if not os.path.exists(local_img):
-                raise FileNotFoundError(f"图片文件不存在: {path}")
-
-            img_pairs.append((idx, local_img))
-
-        # 复用原始处理函数：
-        # 输出文件统一写入 /uploads 目录，文件名带时间戳和随机后缀避免并发冲突
-        base_name = os.path.basename(local_input)
-        name_root, _ = os.path.splitext(base_name)
-        ts = datetime.now().strftime('%Y%m%d%H%M%S')
-        rand = uuid.uuid4().hex[:6]
-        output_filename = f"{name_root}_output_{ts}_{rand}.docx"
-        output_docx = os.path.join(upload_dir, output_filename)
+        parent_dir = os.path.dirname(local_input) or "."
+        fd, output_docx = tempfile.mkstemp(suffix=".docx", dir=parent_dir)
+        os.close(fd)
         process(
             input_docx=local_input,
             output_docx=output_docx,
             replacements=rep_pairs,
-            image_replacements=img_pairs,
             color_keywords=color_keywords,
         )
 
-        # 追加：根据报告类型与标题时间，在“目录”之前自动处理日期和期数
         if report_type or report_title_time:
             try:
                 _apply_report_date_logic_to_docx(
@@ -443,41 +587,95 @@ async def edit_docx_handler(request: Request):
                     report_title_time=report_title_time,
                 )
             except Exception as e:
-                # 避免因为日期处理失败而导致整个接口报错，把错误写到日志即可
                 print(f"apply report date logic failed: {e}")
 
-        abs_out = os.path.abspath(output_docx)
+        os.replace(output_docx, local_input)
+        abs_out = os.path.abspath(local_input)
 
-        # 删除上传的临时文件：只删除位于 tmp 目录中的输入文件
-        try:
-            tmp_root = _get_tmp_upload_dir()
-            if os.path.exists(local_input):
-                abs_input = os.path.abspath(local_input)
-                if os.path.commonpath([abs_input, tmp_root]) == tmp_root:
-                    os.remove(local_input)
-        except Exception:
-            # 不因清理失败影响主流程
-            pass
         return {
-            "output_path": output_docx,
-            "output_url": _build_output_url(output_docx),
+            "output_path": abs_out,
+            "output_url": _build_output_url(abs_out),
         }
-    finally:
-        if tmp_input and os.path.exists(tmp_input):
+    except Exception:
+        if 'output_docx' in locals() and os.path.exists(output_docx):
             try:
-                os.remove(tmp_input)
+                os.remove(output_docx)
             except OSError:
                 pass
+        raise
+    finally:
+        lock_cm.__exit__(None, None, None)
+
+
+@mcp.tool()
+def edit_docx(
+    input_docx_path: str,
+    replacements: List[Dict[str, Any]],
+    report_type: str,
+    report_title_time: str,
+) -> Dict[str, Any]:
+    """
+    对 DOCX 进行文本替换 / 关键字上色。
+
+    支持：
+      - 纯文本替换
+      - 通过 <span color="red">关键字</span> 语法设置关键字颜色
+      - 报告日期与期数自动替换（仅在“目录”之前生效）
+
+    参数:
+      - input_docx_path: 输入文件名
+      - replacements: 替换规则列表，例如
+            [{"old": "原文", "new": "<span color='red'>新文</span>"}]
+      - report_type: 日报/周报/月报 或 daily/weekly/monthly
+      - report_title_time: 替换标题中「YYYY年M月」为指定字符串（首次匹配）
+
+    返回:
+      - 成功: {"success": true, "output_path": ..., "output_url": ...}
+      - 失败: {"success": false, "message": "..."}
+    """
+    try:
+        out = _edit_docx_core(
+            input_docx_path,
+            replacements,
+            report_type,
+            report_title_time,
+        )
+        return {"success": True, **out}
+    except Exception as e:
+        return {"success": False, "message": str(e)}
+
+
+@mcp.custom_route("/edit_docx", methods=["POST"])
+async def edit_docx_handler(request: Request):
+    try:
+        data = await request.json()
+    except json.JSONDecodeError:
+        return JSONResponse(
+            {
+                "success": False,
+                "message": (
+                    "请求体必须是合法的 JSON 对象。"
+                    "请使用 Content-Type: application/json，并发送非空的 JSON body"
+                    "（空 body、form-data 或 urlencoded 会导致此错误）。"
+                ),
+            },
+            status_code=400,
+        )
+    input_docx_path = data.get("input_docx_path")
+    replacements = data.get("replacements")
+    report_type = data.get("report_type")
+    report_title_time = data.get("report_title_time")
+    try:
+        result = _edit_docx_core(
+            input_docx_path,
+            replacements,
+            report_type,
+            report_title_time,
+        )
+        return JSONResponse(result)
+    except Exception as e:
+        return JSONResponse({"success": False, "message": str(e)}, status_code=500)
 
-        for p in tmp_images:
-            if os.path.exists(p):
-                try:
-                    os.remove(p)
-                except OSError:
-                    pass
-# HTTP 远程模式：添加文件上传下载路由
-from starlette.responses import FileResponse, JSONResponse
-from starlette.requests import Request
 def _get_log_path() -> str:
     """
     获取日志文件路径。
@@ -541,28 +739,33 @@ async def upload_handler(request: Request):
         file = form.get("file")
 
         if not file:
-            return JSONResponse({
+                return JSONResponse({
                 "success": False,
                 "message": "未提供文件"
             }, status_code=400)
 
-        tmp_dir = _get_tmp_upload_dir()
+        upload_dir = _get_upload_dir()
         orig_filename = file.filename or "uploaded.docx"
 
         # 安全检查：防止路径遍历攻击，保留原始文件名
-        filename = os.path.basename(orig_filename)
-        file_path = os.path.join(tmp_dir, filename)
+        filename = _safe_filename(orig_filename)
+        file_path = os.path.join(upload_dir, filename)
 
-        # 保存文件到临时目录（如已存在则覆盖）
+        # 保存文件到 uploads 目录（如已存在则覆盖）
         content = await file.read()
-        with open(file_path, "wb") as f:
-            f.write(content)
+        if not content:
+            return JSONResponse({
+                "success": False,
+                "message": f"上传文件为空: {filename}"
+            }, status_code=400)
+        with _file_lock(file_path):
+            _write_bytes_atomic(file_path, content)
 
         return JSONResponse({
             "success": True,
             "filename": filename,        # 保留原始文件名，供 edit_docx 使用
             "file_path": file_path,      # 绝对路径（可选）
-            "file_url": None,            # 临时文件不提供下载 URL
+            "file_url": _build_output_url(file_path),
             "size": len(content),
             "message": f"文件上传成功: {filename}"
         })
@@ -571,33 +774,57 @@ async def upload_handler(request: Request):
             "success": False,
             "message": f"文件上传失败: {str(e)}"
         }, status_code=500)
+@mcp.custom_route("/download", methods=["GET"])
 @mcp.custom_route("/download/{filename}", methods=["GET"])
 async def download_handler(request: Request):
     """处理文件下载"""
     try:
-        filename = request.path_params.get("filename")
+        filename = (
+            request.path_params.get("filename")
+            or request.query_params.get("filename")
+            or request.query_params.get("fileName")
+            or request.query_params.get("name")
+        )
         upload_dir = _get_upload_dir()
+        download_filename = (
+            request.query_params.get("download_filename")
+            or request.query_params.get("new_filename")
+            or request.query_params.get("rename_filename")
+        )
+
+        if not filename:
+            return JSONResponse({
+                "success": False,
+                "message": "缺少 filename 参数"
+            }, status_code=400)
 
         # 安全检查：防止路径遍历攻击
-        filename = os.path.basename(filename)
+        filename = _safe_filename(filename)
         file_path = os.path.join(upload_dir, filename)
-
-        if not os.path.exists(file_path):
-            return JSONResponse({
+        lock_cm = _file_lock(file_path)
+        lock_cm.__enter__()
+        try:
+            if not os.path.exists(file_path):
+                return JSONResponse({
                 "success": False,
                 "message": f"文件不存在: {filename}"
             }, status_code=404)
 
-        if not os.path.isfile(file_path):
-            return JSONResponse({
+            if not os.path.isfile(file_path):
+                return JSONResponse({
                 "success": False,
                 "message": f"不是文件: {filename}"
             }, status_code=400)
 
+            snapshot_path = _snapshot_file(file_path)
+        finally:
+            lock_cm.__exit__(None, None, None)
+
         return FileResponse(
-            file_path,
-            filename=filename,
-            media_type="application/octet-stream"
+            snapshot_path,
+            filename=_safe_filename(download_filename, default=filename),
+            media_type="application/octet-stream",
+            background=BackgroundTask(_cleanup_temp_file, snapshot_path),
         )
     except Exception as e:
         return JSONResponse({