From 21998c07778e87fb3bb1595cf0511dbf690e877e Mon Sep 17 00:00:00 2001
From: hailin <hailin.zhao@gdzx.xyz>
Date: Tue, 7 Apr 2026 02:15:34 -0700
Subject: [PATCH] fix: filter markdown tables, status text, residual formatting
 from antaf

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 modules/antaf/antaf_llm.py | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)

diff --git a/modules/antaf/antaf_llm.py b/modules/antaf/antaf_llm.py
index 78aecf3..c4e606e 100644
--- a/modules/antaf/antaf_llm.py
+++ b/modules/antaf/antaf_llm.py
@@ -40,11 +40,12 @@ class LLMProvider(LLMProviderBase):
 
     @staticmethod
     def _clean_text(text):
-        """清理阿福返回文本中的脏数据、链接、markdown"""
+        """清理阿福返回文本中的脏数据、链接、markdown、表格"""
         import re
         # 去掉阿福内部状态文本
         junk = [
             "完成资料引用", "内容生成", "正在思考", "正在搜索",
+            "开始获取资料", "找到资料",
         ]
         for j in junk:
             text = text.replace(j, "")
@@ -52,12 +53,21 @@ class LLMProvider(LLMProviderBase):
         text = re.sub(r'\[([^\]]+)\]\([^)]+\)', r'\1', text)
         # 裸URL
         text = re.sub(r'https?://\S+', '', text)
-        # Markdown加粗 **文字** → 文字
-        text = re.sub(r'\*\*([^*]+)\*\*', r'\1', text)
+        # Markdown加粗 **文字** → 文字（包括空的 ****）
+        text = re.sub(r'\*{2,}([^*]*)\*{2,}', r'\1', text)
         # Markdown斜体 *文字* → 文字
         text = re.sub(r'\*([^*]+)\*', r'\1', text)
-        # 多余空格
+        # 残留星号
+        text = text.replace('*', '')
+        # Markdown表格行 | xxx | xxx |
+        text = re.sub(r'\|[^|]*\|[^|]*\|[^|\n]*\|?', '', text)
+        # 表格对齐行 | :--- | :--- |
+        text = re.sub(r'\|\s*:?-+:?\s*(\|\s*:?-+:?\s*)+\|?', '', text)
+        # Markdown标题 ### 文字
+        text = re.sub(r'#{1,6}\s*', '', text)
+        # 多余空格和空行
         text = re.sub(r'  +', ' ', text)
+        text = re.sub(r'\n{2,}', '\n', text)
         return text.strip()
 
     @staticmethod