d3f5c82d98
- 新增统一日志工具，支持日志文件路径和级别配置 - 记录 CLI/chat、Agent、LLM、action、MCP、LangGraph、checkpoint 等关键流程 - 对日志中的 token、secret、api_key、Authorization 等敏感信息做脱敏 - chat 新增 llm test 命令，用于验证当前 LLM client 是否正常加载 - 同步 README、打包文档和 run.sh 帮助说明 - 补充日志脱敏和 llm test 相关测试
117 lines
4.5 KiB
Python
117 lines
4.5 KiB
Python
"""Agent 运行日志配置和脱敏工具。"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import json
|
||
import logging
|
||
import os
|
||
import re
|
||
from dataclasses import asdict, is_dataclass
|
||
from pathlib import Path
|
||
from typing import Any
|
||
|
||
from .constants import SENSITIVE_KEYS
|
||
|
||
DEFAULT_LOG_FILE = Path("logs") / "pam_deploy_agent.log"
|
||
LOG_FILE_ENV = "PAM_AGENT_LOG_FILE"
|
||
LOG_LEVEL_ENV = "PAM_AGENT_LOG_LEVEL"
|
||
_HANDLER_MARKER = "_pam_deploy_agent_handler"
|
||
_SENSITIVE_NAME_PARTS = ("secret", "token", "authorization", "api_key", "apikey", "password")
|
||
_ASSIGNMENT_PATTERN = re.compile(
|
||
r"(?i)\b(client_secret|mcp_client_secret|api_key|pam_llm_api_key|token|access_token|authorization|password)\b"
|
||
r"\s*([:=])\s*([^\s,;]+)"
|
||
)
|
||
_AUTH_BEARER_ASSIGNMENT_PATTERN = re.compile(r"(?i)\b(authorization)\b\s*([:=])\s*bearer\s+[^\s,;]+")
|
||
_BEARER_PATTERN = re.compile(r"(?i)(bearer\s+)[A-Za-z0-9._~+\-/=]+")
|
||
|
||
|
||
def configure_logging(
|
||
log_file: str | Path | None = None,
|
||
level: str | int | None = None,
|
||
) -> Path:
|
||
"""配置 Agent 文件日志；重复调用不会重复添加 handler。"""
|
||
actual_path = Path(log_file or os.getenv(LOG_FILE_ENV) or DEFAULT_LOG_FILE)
|
||
actual_path.parent.mkdir(parents=True, exist_ok=True)
|
||
actual_level = _resolve_level(level or os.getenv(LOG_LEVEL_ENV) or "INFO")
|
||
|
||
package_logger = logging.getLogger("pam_deploy_graph")
|
||
package_logger.setLevel(actual_level)
|
||
package_logger.propagate = False
|
||
|
||
marker = str(actual_path.resolve())
|
||
for handler in package_logger.handlers:
|
||
if getattr(handler, _HANDLER_MARKER, "") == marker:
|
||
handler.setLevel(actual_level)
|
||
return actual_path
|
||
|
||
handler = logging.FileHandler(actual_path, encoding="utf-8")
|
||
setattr(handler, _HANDLER_MARKER, marker)
|
||
handler.setLevel(actual_level)
|
||
handler.setFormatter(
|
||
logging.Formatter(
|
||
fmt="%(asctime)s %(levelname)s [%(name)s] %(message)s",
|
||
datefmt="%Y-%m-%d %H:%M:%S",
|
||
)
|
||
)
|
||
package_logger.addHandler(handler)
|
||
package_logger.info("日志已初始化 path=%s level=%s", actual_path, logging.getLevelName(actual_level))
|
||
return actual_path
|
||
|
||
|
||
def redact_for_log(value: Any, *, max_text_len: int = 1200) -> Any:
|
||
"""递归脱敏并截断日志对象，避免把密钥、token 或完整长文本写入日志。"""
|
||
if is_dataclass(value) and not isinstance(value, type):
|
||
return redact_for_log(asdict(value), max_text_len=max_text_len)
|
||
if isinstance(value, dict):
|
||
redacted: dict[str, Any] = {}
|
||
for key, item in value.items():
|
||
text_key = str(key)
|
||
if _is_sensitive_key(text_key):
|
||
redacted[text_key] = "***"
|
||
else:
|
||
redacted[text_key] = redact_for_log(item, max_text_len=max_text_len)
|
||
return redacted
|
||
if isinstance(value, (list, tuple, set)):
|
||
return [redact_for_log(item, max_text_len=max_text_len) for item in value]
|
||
if isinstance(value, str):
|
||
return _truncate(_redact_string(value), max_text_len)
|
||
if value is None or isinstance(value, (bool, int, float)):
|
||
return value
|
||
return _truncate(_redact_string(str(value)), max_text_len)
|
||
|
||
|
||
def json_for_log(value: Any, *, max_text_len: int = 1200) -> str:
|
||
"""把对象脱敏后序列化成适合单行日志的 JSON 文本。"""
|
||
redacted = redact_for_log(value, max_text_len=max_text_len)
|
||
return json.dumps(redacted, ensure_ascii=False, default=str, sort_keys=True)
|
||
|
||
|
||
def _resolve_level(value: str | int) -> int:
|
||
"""解析日志级别字符串，非法值降级为 INFO。"""
|
||
if isinstance(value, int):
|
||
return value
|
||
resolved = getattr(logging, str(value).upper(), logging.INFO)
|
||
return resolved if isinstance(resolved, int) else logging.INFO
|
||
|
||
|
||
def _is_sensitive_key(key: str) -> bool:
|
||
"""判断字段名是否应脱敏。"""
|
||
if key in SENSITIVE_KEYS:
|
||
return True
|
||
normalized = key.lower().replace("-", "_")
|
||
return any(part in normalized for part in _SENSITIVE_NAME_PARTS)
|
||
|
||
|
||
def _truncate(value: str, limit: int) -> str:
|
||
"""截断过长字符串。"""
|
||
if len(value) <= limit:
|
||
return value
|
||
return value[:limit] + "...[已截断]"
|
||
|
||
|
||
def _redact_string(value: str) -> str:
|
||
"""脱敏字符串中的常见 KEY=VALUE 和 Bearer token 片段。"""
|
||
value = _AUTH_BEARER_ASSIGNMENT_PATTERN.sub(lambda match: f"{match.group(1)}{match.group(2)}***", value)
|
||
value = _ASSIGNMENT_PATTERN.sub(lambda match: f"{match.group(1)}{match.group(2)}***", value)
|
||
return _BEARER_PATTERN.sub(lambda match: f"{match.group(1)}***", value)
|