fix(security): 代码审查安全加固 — 三批次修复(6H/13M/12L)

第一批(快速修复):
- H6: main.py --debug 默认值 True→False,防止 Werkzeug RCE
- M2+M3: /search 增加 validate_query + top_k 范围限制(1-50)
- M4: context_count 范围限制(0-10) + 异常捕获
- L3: assert → raise RuntimeError(生产环境 API Key 检查)
- H1: SSE 错误事件移除 traceback 字段

第二批(安全加固):
- H2+H3: 文档接口路径遍历 realpath 校验 + 文件类型/大小限制
- H4+H5: 批量上传文件大小检查
- M6: LIKE 查询通配符转义
- M1: 37 处 str(e) 异常信息统一脱敏(6 文件)
- M5: CORS 生产环境限制来源
- M7: SESSION_MANAGER None 保护(503)
- M11: subprocess 参数注入防护(白名单 + -- 分隔符)

第三批(架构改进):
- M8+M9: 提取 JSON 解析共享工具(extract_json_object/list)
- M10: Prompt 注入检测防御(prompt_guard.py)
- M12: 解析器文件大小限制(Excel 50MB/TXT 20MB/PDF 100MB)
- M13: 全局单例竞态条件双重检查锁定(engine/bm25/intent_analyzer)
This commit is contained in:
lacerate551
2026-06-05 15:26:32 +08:00
parent a7206377cc
commit 90b915232a
19 changed files with 436 additions and 102 deletions

View File

@@ -391,10 +391,11 @@ def sync_documents() -> Tuple[Any, int]:
}
})
except Exception as e:
logger.error(f"知识库操作异常: {e}")
results.append({
"collection": "all",
"status": "error",
"message": str(e)
"message": "操作失败"
})
else:
# 没有 sync_service返回提示
@@ -462,7 +463,8 @@ def debug_scan() -> Tuple[Any, int]:
result["scanned_count"] = len(scanned)
result["scanned_ids"] = list(scanned.keys())[:10]
except Exception as e:
result["scan_error"] = str(e)
logger.error(f"扫描异常: {e}")
result["scan_error"] = "扫描失败"
return jsonify(result)
@@ -499,9 +501,11 @@ def reindex_collection(kb_name: str) -> Tuple[Any, int]:
from data.db import get_connection
with get_connection("knowledge") as conn:
cursor = conn.cursor()
# 转义 LIKE 通配符,防止 kb_name 中的 % 或 _ 导致非预期匹配
escaped_kb = kb_name.replace('%', '\\%').replace('_', '\\_')
# 删除以 "{kb_name}/" 或 "{kb_name}\" 开头的文档哈希(兼容 Windows 和 Linux
cursor.execute("DELETE FROM document_hashes WHERE document_id LIKE ? OR document_id LIKE ?",
(f"{kb_name}/%", f"{kb_name}\\%"))
cursor.execute("DELETE FROM document_hashes WHERE document_id LIKE ? ESCAPE '\\' OR document_id LIKE ? ESCAPE '\\'",
(f"{escaped_kb}/%", f"{escaped_kb}\\%"))
deleted = cursor.rowcount
logger.info(f"已清除 {deleted} 条哈希记录: {kb_name}")
except Exception as e:
@@ -520,7 +524,8 @@ def reindex_collection(kb_name: str) -> Tuple[Any, int]:
"errors": result.errors
})
except Exception as e:
return jsonify({"error": str(e)}), 500
logger.error(f"reindex 异常: {e}")
return jsonify({"error": "操作失败,请稍后重试"}), 500
else:
return jsonify({"error": "同步服务不可用"}), 503
@@ -633,7 +638,8 @@ def deprecate_document(kb_name: str, filename: str) -> Tuple[Any, int]:
)
return jsonify(result)
except Exception as e:
return jsonify({"success": False, "error": str(e)}), 500
logger.error(f"操作异常: {e}")
return jsonify({"success": False, "error": "操作失败,请稍后重试"}), 500
@kb_bp.route('/collections/<kb_name>/documents/<path:filename>/restore', methods=['POST'])
@@ -664,7 +670,8 @@ def restore_document(kb_name: str, filename: str) -> Tuple[Any, int]:
result = kb_manager.restore_document(kb_name, filename)
return jsonify(result)
except Exception as e:
return jsonify({"success": False, "error": str(e)}), 500
logger.error(f"操作异常: {e}")
return jsonify({"success": False, "error": "操作失败,请稍后重试"}), 500
@kb_bp.route('/collections/<kb_name>/documents/<path:filename>/versions', methods=['GET'])
@@ -720,7 +727,8 @@ def get_document_versions(kb_name: str, filename: str) -> Tuple[Any, int]:
"total": len(versions_data)
})
except Exception as e:
return jsonify({"success": False, "error": str(e)}), 500
logger.error(f"操作异常: {e}")
return jsonify({"success": False, "error": "操作失败,请稍后重试"}), 500
@kb_bp.route('/collections/<kb_name>/update-image-descriptions', methods=['POST'])
@@ -754,7 +762,8 @@ def update_image_descriptions(kb_name: str) -> Tuple[Any, int]:
result = kb_manager.update_image_descriptions(kb_name)
return jsonify(result)
except Exception as e:
return jsonify({"success": False, "error": str(e)}), 500
logger.error(f"操作异常: {e}")
return jsonify({"success": False, "error": "操作失败,请稍后重试"}), 500
@kb_bp.route('/collections/sync-vlm-cache', methods=['POST'])