修复markdown出错时不返回原文的问题，优化提示词

docutranslate/agents/glossary_agent.py

@@ -51,7 +51,7 @@ Output:
 
 
 def get_original_segments(prompt: str):
-    match = re.search(r'<input>(.*)</input>', prompt, re.DOTALL)
+    match = re.search(r'<input>\n```json\n(.*)\n```\n</input>', prompt, re.DOTALL)
     if match:
         return match.group(1)
     else:

docutranslate/agents/markdown_agent.py

@@ -1,11 +1,20 @@
 # SPDX-FileCopyrightText: 2025 QinHan
 # SPDX-License-Identifier: MPL-2.0
-
+import re
 from dataclasses import dataclass
 
 from .agent import Agent, AgentConfig
 from ..glossary.glossary import Glossary
 
+
+def get_original_markdown(prompt: str):
+    match = re.search(r'<input>\n(.*)\n</input>', prompt, re.DOTALL)
+    if match:
+        return match.group(1)
+    else:
+        raise ValueError("无法从prompt中提取初始文本")
+
+
 def generate_prompt(markdown_text: str, to_lang: str):
     return f"""
 Treat the text input as markdown text and translate it into {to_lang},output translation ONLY. 
@@ -20,9 +29,12 @@ Treat the text input as markdown text and translate it into {to_lang},output tra
 - Output the translated markdown text as plain text (not in a markdown code block, with no extraneous text).
 
 The markdown text input:
+<input>
  {markdown_text}
+</input>
 """
 
+
 @dataclass
 class MDTranslateAgentConfig(AgentConfig):
     to_lang: str
@@ -51,11 +63,14 @@ You are a professional machine translation engine.
 
     def send_chunks(self, prompts: list[str]):
         prompts = [generate_prompt(prompt, self.to_lang) for prompt in prompts]
-        return super().send_prompts(prompts=prompts, pre_send_handler=self._pre_send_handler)
+        return super().send_prompts(prompts=prompts, pre_send_handler=self._pre_send_handler,
+                                    error_result_handler=lambda prompt, logger: get_original_markdown(prompt))
 
     async def send_chunks_async(self, prompts: list[str]):
         prompts = [generate_prompt(prompt, self.to_lang) for prompt in prompts]
-        return await super().send_prompts_async(prompts=prompts, pre_send_handler=self._pre_send_handler)
+        return await super().send_prompts_async(prompts=prompts, pre_send_handler=self._pre_send_handler,
+                                                error_result_handler=lambda prompt, logger: get_original_markdown(
+                                                    prompt))
 
     def update_glossary_dict(self, update_dict: dict | None):
         if self.glossary_dict is None:

docutranslate/agents/segments_agent.py

@@ -22,11 +22,9 @@ You will receive a sequence of original text segments to be translated, represen
 Here is the input:
 
 <input>
-
 ```json
 {json_segments}
 ```
-
 </input>
 
 For each Key-Value Pair in the JSON, translate the contents of the value into {to_lang}, Write the translation back into the value for that JSON.
@@ -60,7 +58,7 @@ Please return the translated JSON directly without including any additional info
 """
 
 def get_original_segments(prompt:str):
-    match = re.search(r'<input>(.*)</input>', prompt, re.DOTALL)
+    match = re.search(r'<input>\n```json\n(.*)\n```\n</input>', prompt, re.DOTALL)
     if match:
         return match.group(1)
     else:

docutranslate/cacher/md_based_convert_cacher.py

@@ -27,14 +27,18 @@ class MDBasedCovertCacher:
 
     def get_cached_result(self, document: Document, convert_engin: str,
                           convert_config: ConverterConfig) -> MarkdownDocument | None:
-        return self.cache_dict.get(self._get_hashcode(document, convert_engin, convert_config))
+        d: MarkdownDocument | None = self.cache_dict.get(self._get_hashcode(document, convert_engin, convert_config))
+        if d:
+            return d.copy()
+        else:
+            return None
 
     def cache_result(self, convert_result: MarkdownDocument, document: Document, convert_engin: str,
                      convert_config: ConverterConfig) -> MarkdownDocument:
         hash_code = self._get_hashcode(document, convert_engin, convert_config)
         if len(self.cache_dict) > int(CACHE_NUM):
             self.cache_dict.popitem(last=False)
-        self.cache_dict[hash_code] = convert_result
+        self.cache_dict[hash_code] = convert_result.copy()
         return convert_result
 
     def clear(self):

docutranslate/workflow/md_based_workflow.py

@@ -83,7 +83,7 @@ class MarkdownBasedWorkflow(Workflow[MarkdownBasedWorkflowConfig, Document, Mark
             for attachment in converter.attachments:
                 self.attachment.add_attachment(attachment)
         # 缓存解析后文件
-        md_based_convert_cacher.cache_result(document_md.copy(), self.document_original, convert_engin, convert_config)
+        md_based_convert_cacher.cache_result(document_md, self.document_original, convert_engin, convert_config)
 
         return document_md