From bf1e2bfa9c1a18a0ddd3b3815991593e42c52835 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E9=A9=AC=E4=B8=80=E4=B8=81?= <1769123563@qq.com>
Date: Mon, 17 Nov 2025 22:10:37 +0800
Subject: [PATCH] Repair the Logic for Cleaning Data Returned by LLM
---
ReportEngine/utils/json_parser.py | 33 ++++++++++++++++++-------------
1 file changed, 19 insertions(+), 14 deletions(-)
diff --git a/ReportEngine/utils/json_parser.py b/ReportEngine/utils/json_parser.py
index 1de6a19..0184591 100644
--- a/ReportEngine/utils/json_parser.py
+++ b/ReportEngine/utils/json_parser.py
@@ -51,12 +51,12 @@ class RobustJSONParser:
# 常见的LLM思考内容模式
_THINKING_PATTERNS = [
- r".*?",
- r".*?",
- r"让我想想.*?(?=\{|\[|$)",
- r"首先.*?(?=\{|\[|$)",
- r"分析.*?(?=\{|\[|$)",
- r"根据.*?(?=\{|\[|$)",
+ r"^\s*.*?\s*",
+ r"^\s*.*?\s*",
+ r"^\s*让我想想.*?(?=\{|\[|$)",
+ r"^\s*首先.*?(?=\{|\[|$)",
+ r"^\s*分析.*?(?=\{|\[|$)",
+ r"^\s*根据.*?(?=\{|\[|$)",
]
# 冒号等号模式（LLM常见错误）
@@ -182,16 +182,21 @@ class RobustJSONParser:
for pattern in self._THINKING_PATTERNS:
cleaned = re.sub(pattern, "", cleaned, flags=re.DOTALL | re.IGNORECASE)
- # 移除markdown代码块标记
- if cleaned.startswith("```json"):
- cleaned = cleaned[7:]
- elif cleaned.startswith("```"):
- cleaned = cleaned[3:]
+ # 优先提取任意位置的```json```包裹内容
+ fenced_match = re.search(r"```(?:json)?\s*([\s\S]*?)\s*```", cleaned)
+ if fenced_match:
+ cleaned = fenced_match.group(1).strip()
+ else:
+ # 如果没有找到完整代码块，再尝试移除前后缀
+ if cleaned.startswith("```json"):
+ cleaned = cleaned[7:]
+ elif cleaned.startswith("```"):
+ cleaned = cleaned[3:]
- if cleaned.endswith("```"):
- cleaned = cleaned[:-3]
+ if cleaned.endswith("```"):
+ cleaned = cleaned[:-3]
- cleaned = cleaned.strip()
+ cleaned = cleaned.strip()
# 尝试提取第一个完整的JSON对象或数组
cleaned = self._extract_first_json_structure(cleaned)