agent-Specialization/modules/personalization_manager.py
JOJO 72a49a7c8d fix: 修复对话计时器持久化、子智能体工具美化与状态查询,统一后台通知池
- 新增 server/work_timer.py,在对话真正空闲(无前台任务、无后台子智能体/后台命令/压缩)时持久化 work_timer,并同步内存副本,解决刷新后计时器回退问题。
- server/chat_flow_task_main.py / chat_flow.py / tasks/models.py 在任务结束/取消时按空闲判定决定是否持久化。
- 前端 lifecycle.ts 在仍有后台任务时不提前停止计时器。
- 子智能体工具(create/terminate/get_status)渲染美化,task 参数放在顶部元信息区。
- 子智能体状态查询支持返回「已完成」「已终止」「不存在」;修复 wait_for_completion 在 final_result 就绪前返回导致的「ID 被占用」误报。
- 统一后台完成通知池 poll_completion_notifications,合并子智能体与后台 run_command 两路轮询,避免逐条触发工作循环与单工作区互斥冲突。
- 删除本次新增的各类 debug_log / notify_pool_log 调用及辅助脚本。
2026-06-25 04:16:44 +08:00

798 lines
33 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

"""Utilities for managing per-user personalization settings."""
from __future__ import annotations
import json
from copy import deepcopy
from pathlib import Path
from typing import Any, Dict, Iterable, Optional, Union
try:
from config.limits import THINKING_FAST_INTERVAL
except ImportError:
THINKING_FAST_INTERVAL = 10
from core.tool_config import TOOL_CATEGORIES
from config.model_profiles import get_default_model_key, get_registered_model_keys
ALLOWED_RUN_MODES = {"fast", "thinking", "deep"}
ALLOWED_PERMISSION_MODES = {"readonly", "approval", "auto_approval", "unrestricted"}
ALLOWED_THEMES = {"classic", "light", "dark"}
ALLOWED_COMMUNICATION_STYLES = {"default", "human_like", "auto"}
ALLOWED_CONVERSATION_CONTINUITY = {"low", "medium", "high"}
ALLOWED_GOAL_REVIEW_MODES = {"readonly", "active"}
ALLOWED_GOAL_END_CONDITIONS = {"max_turns", "max_tokens"}
GOAL_MAX_TURNS_MIN = 1
GOAL_MAX_TURNS_MAX = 100
GOAL_MAX_TURNS_DEFAULT = 5
GOAL_MAX_TOKENS_MIN = 1_000
GOAL_MAX_TOKENS_MAX = 100_000_000
PERSONALIZATION_FILENAME = "personalization.json"
MAX_SHORT_FIELD_LENGTH = 20
MAX_CONSIDERATION_LENGTH = 50
MAX_CONSIDERATION_ITEMS = 10
TONE_PRESETS = ["健谈", "幽默", "直言不讳", "鼓励性", "诗意", "企业商务", "打破常规", "同理心"]
THINKING_INTERVAL_MIN = 1
THINKING_INTERVAL_MAX = 50
RECENT_CONVERSATIONS_PROMPT_LIMIT_MIN = 1
RECENT_CONVERSATIONS_PROMPT_LIMIT_MAX = 30
RECENT_CONVERSATIONS_PROMPT_LIMIT_DEFAULT = 10
DEFAULT_SHALLOW_COMPRESS_TRIGGER_TOKENS = 80_000
DEFAULT_SHALLOW_COMPRESS_KEEP_RECENT_TOOLS = 15
DEFAULT_SHALLOW_COMPRESS_MAX_REPLACE_PER_ROUND = 10
DEFAULT_SHALLOW_COMPRESS_TRIGGER_TOOL_CALLS_INTERVAL = 10
DEFAULT_DEEP_COMPRESS_TRIGGER_TOKENS = 150_000
MIN_COMPRESSION_TRIGGER_TOKENS = 1_000
MAX_COMPRESSION_TRIGGER_TOKENS = 2_000_000
MIN_SHALLOW_KEEP_RECENT_TOOLS = 0
MAX_SHALLOW_KEEP_RECENT_TOOLS = 500
MIN_SHALLOW_MAX_REPLACE_PER_ROUND = 1
MAX_SHALLOW_MAX_REPLACE_PER_ROUND = 200
MIN_SHALLOW_TRIGGER_TOOL_CALLS_INTERVAL = 1
MAX_SHALLOW_TRIGGER_TOOL_CALLS_INTERVAL = 200
MIN_SHALLOW_KEEP_USER_TURN_TOOLS = 0
MAX_SHALLOW_KEEP_USER_TURN_TOOLS = 50
DEFAULT_SHALLOW_KEEP_USER_TURN_TOOLS = 3
DEFAULT_PERSONALIZATION_CONFIG: Dict[str, Any] = {
"enabled": False,
"communication_style": "default", # default / human_like / auto
"conversation_continuity": "medium", # high / medium / low
"self_identify": "",
"user_name": "",
"use_custom_names": False,
"profession": "",
"tone": "",
"considerations": [],
"thinking_interval": None,
"disabled_tool_categories": [],
"enabled_skills": None,
"skills_catalog_snapshot": None,
"default_run_mode": "deep",
"default_permission_mode": "unrestricted",
"auto_generate_title": True,
"recent_conversations_prompt_enabled": False,
"recent_conversations_prompt_limit": RECENT_CONVERSATIONS_PROMPT_LIMIT_DEFAULT,
"tool_intent_enabled": True,
"skill_hints_enabled": False, # Skill 提示系统开关(默认关闭)
"skill_strict_terminal_enabled": False, # 强约束terminal 系列工具需先阅读 terminal-guide
"skill_strict_sub_agent_enabled": False, # 强约束:子智能体系列工具需先阅读 sub-agent-guide
"skill_strict_run_command_foreground_enabled": False, # 强约束run_command 前台模式需先阅读 run-command-guide
"skill_strict_run_command_background_enabled": False, # 强约束run_command 后台模式需先阅读 run-command-guide
"default_model": None,
"image_compression": "original", # original / 1080p / 720p / 540p
"auto_shallow_compress_enabled": False,
"auto_deep_compress_enabled": True,
"shallow_compress_trigger_tokens": None,
"shallow_compress_keep_recent_tools": None,
"shallow_compress_max_replace_per_round": None,
"shallow_compress_trigger_tool_calls_interval": None,
"shallow_compress_keep_user_turn_tools": None, # 保留最近N次用户输入后的工具不压缩默认3
"deep_compress_trigger_tokens": None,
"deep_compress_form": "file", # 深压缩形式file-生成文件 / inject-直接注入文件全文
"silent_tool_disable": True, # 禁用工具时不向模型插入提示(默认开启)
"enhanced_tool_display": True, # 增强工具显示
"compact_message_display": "full", # 简略消息显示full-完整原始内容 / brief-一行概要
"stacked_hide_borders": False, # 堆叠块隐藏边线
"show_git_status_bar": True, # 是否显示输入栏上方 Git 状态栏
"versioning_restore_mode": "overwrite", # 版本回溯模式固定为 overwrite
"agents_md_auto_inject": False, # AGENTS.md 自动注入开关
"allow_root_file_creation": False, # 允许在根目录创建文件开关
"default_hide_workspace": False, # 默认隐藏工作区
"theme": "classic", # 主题配色: classic-经典/light-明亮/dark-暗黑
# 目标模式Goal Mode
"goal_review_mode": "readonly", # readonly-仅读对话判断 / active-允许审核智能体跑只读命令取证
"goal_end_conditions": ["max_turns"], # 结束方式可多选max_turns / max_tokens
"goal_max_turns": GOAL_MAX_TURNS_DEFAULT, # 最多自动续命轮数
"goal_max_tokens": None, # 累计(输入+输出)token 上限None 表示不启用
}
__all__ = [
"PERSONALIZATION_FILENAME",
"DEFAULT_PERSONALIZATION_CONFIG",
"TONE_PRESETS",
"MAX_CONSIDERATION_ITEMS",
"RECENT_CONVERSATIONS_PROMPT_LIMIT_MIN",
"RECENT_CONVERSATIONS_PROMPT_LIMIT_MAX",
"RECENT_CONVERSATIONS_PROMPT_LIMIT_DEFAULT",
"load_personalization_config",
"save_personalization_config",
"ensure_personalization_config",
"build_personalization_prompt",
"sanitize_personalization_payload",
"resolve_context_compression_settings",
"validate_context_compression_settings",
"ALLOWED_PERMISSION_MODES",
]
PathLike = Union[str, Path]
def _ensure_parent(path: Path) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
def _to_path(base: PathLike) -> Path:
base_path = Path(base).expanduser()
if base_path.is_dir():
return base_path / PERSONALIZATION_FILENAME
return base_path
def ensure_personalization_config(base_dir: PathLike) -> Dict[str, Any]:
"""Ensure the personalization file exists and return its content."""
path = _to_path(base_dir)
_ensure_parent(path)
if not path.exists():
with open(path, "w", encoding="utf-8") as f:
json.dump(DEFAULT_PERSONALIZATION_CONFIG, f, ensure_ascii=False, indent=2)
return deepcopy(DEFAULT_PERSONALIZATION_CONFIG)
return load_personalization_config(base_dir)
def load_personalization_config(base_dir: PathLike) -> Dict[str, Any]:
"""Load personalization config; fall back to defaults on errors."""
path = _to_path(base_dir)
_ensure_parent(path)
if not path.exists():
return ensure_personalization_config(base_dir)
try:
with open(path, "r", encoding="utf-8") as f:
raw = json.load(f)
sanitized = sanitize_personalization_payload(raw)
# 若发现缺失字段(如默认模型)或数据被规范化,主动写回文件,避免下一次读取仍为旧格式
if sanitized != raw:
with open(path, "w", encoding="utf-8") as wf:
json.dump(sanitized, wf, ensure_ascii=False, indent=2)
return sanitized
except (json.JSONDecodeError, OSError):
# 重置为默认配置,避免错误阻塞
with open(path, "w", encoding="utf-8") as f:
json.dump(DEFAULT_PERSONALIZATION_CONFIG, f, ensure_ascii=False, indent=2)
return deepcopy(DEFAULT_PERSONALIZATION_CONFIG)
def sanitize_personalization_payload(
payload: Optional[Dict[str, Any]],
fallback: Optional[Dict[str, Any]] = None
) -> Dict[str, Any]:
"""Normalize payload structure and clamp field lengths."""
base = deepcopy(DEFAULT_PERSONALIZATION_CONFIG)
if fallback:
base.update(fallback)
data = payload or {}
allowed_tool_categories = set(TOOL_CATEGORIES.keys())
allowed_models = set(get_registered_model_keys(visible_only=True))
allowed_image_modes = {"original", "1080p", "720p", "540p"}
def _resolve_short_field(key: str) -> str:
if key in data:
return _sanitize_short_field(data.get(key))
return _sanitize_short_field(base.get(key))
base["enabled"] = bool(data.get("enabled", base["enabled"]))
# 交流风格: default / human_like / auto
_comm_style = data.get("communication_style", base.get("communication_style", "default"))
base["communication_style"] = _comm_style if _comm_style in ALLOWED_COMMUNICATION_STYLES else "default"
_conversation_continuity = data.get(
"conversation_continuity",
base.get("conversation_continuity", "medium"),
)
base["conversation_continuity"] = (
_conversation_continuity
if _conversation_continuity in ALLOWED_CONVERSATION_CONTINUITY
else "medium"
)
base["auto_generate_title"] = bool(data.get("auto_generate_title", base["auto_generate_title"]))
base["recent_conversations_prompt_enabled"] = bool(
data.get("recent_conversations_prompt_enabled", base.get("recent_conversations_prompt_enabled", False))
)
base["recent_conversations_prompt_limit"] = (
_sanitize_optional_int(
data.get("recent_conversations_prompt_limit", base.get("recent_conversations_prompt_limit")),
min_value=RECENT_CONVERSATIONS_PROMPT_LIMIT_MIN,
max_value=RECENT_CONVERSATIONS_PROMPT_LIMIT_MAX,
)
or RECENT_CONVERSATIONS_PROMPT_LIMIT_DEFAULT
)
base["self_identify"] = _resolve_short_field("self_identify")
base["user_name"] = _resolve_short_field("user_name")
base["profession"] = _resolve_short_field("profession")
base["tone"] = _resolve_short_field("tone")
if "considerations" in data:
base["considerations"] = _sanitize_considerations(data.get("considerations"))
else:
base["considerations"] = _sanitize_considerations(base.get("considerations", []))
if "thinking_interval" in data:
base["thinking_interval"] = _sanitize_thinking_interval(data.get("thinking_interval"))
else:
base["thinking_interval"] = _sanitize_thinking_interval(base.get("thinking_interval"))
# 工具意图提示开关
if "tool_intent_enabled" in data:
base["tool_intent_enabled"] = bool(data.get("tool_intent_enabled"))
else:
base["tool_intent_enabled"] = bool(base.get("tool_intent_enabled"))
# Skill 提示系统开关
if "skill_hints_enabled" in data:
base["skill_hints_enabled"] = bool(data.get("skill_hints_enabled"))
else:
base["skill_hints_enabled"] = bool(base.get("skill_hints_enabled"))
# Skill 强约束开关(仅针对已启用的 skill 生效)
if "skill_strict_terminal_enabled" in data:
base["skill_strict_terminal_enabled"] = bool(data.get("skill_strict_terminal_enabled"))
else:
base["skill_strict_terminal_enabled"] = bool(base.get("skill_strict_terminal_enabled", False))
if "skill_strict_sub_agent_enabled" in data:
base["skill_strict_sub_agent_enabled"] = bool(data.get("skill_strict_sub_agent_enabled"))
else:
base["skill_strict_sub_agent_enabled"] = bool(base.get("skill_strict_sub_agent_enabled", False))
if "skill_strict_run_command_foreground_enabled" in data:
base["skill_strict_run_command_foreground_enabled"] = bool(data.get("skill_strict_run_command_foreground_enabled"))
else:
base["skill_strict_run_command_foreground_enabled"] = bool(
base.get("skill_strict_run_command_foreground_enabled", False)
)
if "skill_strict_run_command_background_enabled" in data:
base["skill_strict_run_command_background_enabled"] = bool(data.get("skill_strict_run_command_background_enabled"))
else:
base["skill_strict_run_command_background_enabled"] = bool(
base.get("skill_strict_run_command_background_enabled", False)
)
if "disabled_tool_categories" in data:
base["disabled_tool_categories"] = _sanitize_tool_categories(data.get("disabled_tool_categories"), allowed_tool_categories)
else:
base["disabled_tool_categories"] = _sanitize_tool_categories(base.get("disabled_tool_categories"), allowed_tool_categories)
if "enabled_skills" in data:
base["enabled_skills"] = _sanitize_skills(data.get("enabled_skills"))
else:
base["enabled_skills"] = _sanitize_skills(base.get("enabled_skills"))
if "skills_catalog_snapshot" in data:
base["skills_catalog_snapshot"] = _sanitize_skills(data.get("skills_catalog_snapshot"))
else:
base["skills_catalog_snapshot"] = _sanitize_skills(base.get("skills_catalog_snapshot"))
if "default_run_mode" in data:
base["default_run_mode"] = _sanitize_run_mode(data.get("default_run_mode"))
else:
base["default_run_mode"] = _sanitize_run_mode(base.get("default_run_mode"))
permission_mode = data.get("default_permission_mode", base.get("default_permission_mode"))
if isinstance(permission_mode, str) and permission_mode in ALLOWED_PERMISSION_MODES:
base["default_permission_mode"] = permission_mode
elif base.get("default_permission_mode") not in ALLOWED_PERMISSION_MODES:
base["default_permission_mode"] = "unrestricted"
base["versioning_restore_mode"] = "overwrite"
# 默认模型
chosen_model = data.get("default_model", base.get("default_model"))
if isinstance(chosen_model, str) and chosen_model in allowed_models:
base["default_model"] = chosen_model
elif allowed_models and base.get("default_model") not in allowed_models:
try:
base["default_model"] = get_default_model_key(visible_only=True)
except Exception:
base["default_model"] = None
# 图片压缩模式
img_mode = data.get("image_compression", base.get("image_compression"))
if isinstance(img_mode, str) and img_mode in allowed_image_modes:
base["image_compression"] = img_mode
elif base.get("image_compression") not in allowed_image_modes:
base["image_compression"] = "original"
if "auto_shallow_compress_enabled" in data:
base["auto_shallow_compress_enabled"] = bool(data.get("auto_shallow_compress_enabled"))
else:
base["auto_shallow_compress_enabled"] = bool(base.get("auto_shallow_compress_enabled", False))
if "auto_deep_compress_enabled" in data:
base["auto_deep_compress_enabled"] = bool(data.get("auto_deep_compress_enabled"))
else:
base["auto_deep_compress_enabled"] = bool(base.get("auto_deep_compress_enabled", False))
if "shallow_compress_trigger_tokens" in data:
base["shallow_compress_trigger_tokens"] = _sanitize_optional_int(
data.get("shallow_compress_trigger_tokens"),
min_value=MIN_COMPRESSION_TRIGGER_TOKENS,
max_value=MAX_COMPRESSION_TRIGGER_TOKENS,
)
else:
base["shallow_compress_trigger_tokens"] = _sanitize_optional_int(
base.get("shallow_compress_trigger_tokens"),
min_value=MIN_COMPRESSION_TRIGGER_TOKENS,
max_value=MAX_COMPRESSION_TRIGGER_TOKENS,
)
if "shallow_compress_keep_recent_tools" in data:
base["shallow_compress_keep_recent_tools"] = _sanitize_optional_int(
data.get("shallow_compress_keep_recent_tools"),
min_value=MIN_SHALLOW_KEEP_RECENT_TOOLS,
max_value=MAX_SHALLOW_KEEP_RECENT_TOOLS,
)
else:
base["shallow_compress_keep_recent_tools"] = _sanitize_optional_int(
base.get("shallow_compress_keep_recent_tools"),
min_value=MIN_SHALLOW_KEEP_RECENT_TOOLS,
max_value=MAX_SHALLOW_KEEP_RECENT_TOOLS,
)
if "shallow_compress_keep_user_turn_tools" in data:
base["shallow_compress_keep_user_turn_tools"] = _sanitize_optional_int(
data.get("shallow_compress_keep_user_turn_tools"),
min_value=MIN_SHALLOW_KEEP_USER_TURN_TOOLS,
max_value=MAX_SHALLOW_KEEP_USER_TURN_TOOLS,
)
else:
base["shallow_compress_keep_user_turn_tools"] = _sanitize_optional_int(
base.get("shallow_compress_keep_user_turn_tools"),
min_value=MIN_SHALLOW_KEEP_USER_TURN_TOOLS,
max_value=MAX_SHALLOW_KEEP_USER_TURN_TOOLS,
)
if "shallow_compress_max_replace_per_round" in data:
base["shallow_compress_max_replace_per_round"] = _sanitize_optional_int(
data.get("shallow_compress_max_replace_per_round"),
min_value=MIN_SHALLOW_MAX_REPLACE_PER_ROUND,
max_value=MAX_SHALLOW_MAX_REPLACE_PER_ROUND,
)
else:
base["shallow_compress_max_replace_per_round"] = _sanitize_optional_int(
base.get("shallow_compress_max_replace_per_round"),
min_value=MIN_SHALLOW_MAX_REPLACE_PER_ROUND,
max_value=MAX_SHALLOW_MAX_REPLACE_PER_ROUND,
)
if "deep_compress_trigger_tokens" in data:
base["deep_compress_trigger_tokens"] = _sanitize_optional_int(
data.get("deep_compress_trigger_tokens"),
min_value=MIN_COMPRESSION_TRIGGER_TOKENS,
max_value=MAX_COMPRESSION_TRIGGER_TOKENS,
)
else:
base["deep_compress_trigger_tokens"] = _sanitize_optional_int(
base.get("deep_compress_trigger_tokens"),
min_value=MIN_COMPRESSION_TRIGGER_TOKENS,
max_value=MAX_COMPRESSION_TRIGGER_TOKENS,
)
# 深压缩形式file生成文件引导语提示位置/ inject直接注入文件全文
deep_form = data.get("deep_compress_form", base.get("deep_compress_form"))
if isinstance(deep_form, str) and deep_form.strip().lower() in ("file", "inject"):
base["deep_compress_form"] = deep_form.strip().lower()
else:
base["deep_compress_form"] = "file"
if "shallow_compress_trigger_tool_calls_interval" in data:
base["shallow_compress_trigger_tool_calls_interval"] = _sanitize_optional_int(
data.get("shallow_compress_trigger_tool_calls_interval"),
min_value=MIN_SHALLOW_TRIGGER_TOOL_CALLS_INTERVAL,
max_value=MAX_SHALLOW_TRIGGER_TOOL_CALLS_INTERVAL,
)
else:
base["shallow_compress_trigger_tool_calls_interval"] = _sanitize_optional_int(
base.get("shallow_compress_trigger_tool_calls_interval"),
min_value=MIN_SHALLOW_TRIGGER_TOOL_CALLS_INTERVAL,
max_value=MAX_SHALLOW_TRIGGER_TOOL_CALLS_INTERVAL,
)
# 静默禁用工具提示
if "silent_tool_disable" in data:
base["silent_tool_disable"] = bool(data.get("silent_tool_disable"))
else:
base["silent_tool_disable"] = bool(base.get("silent_tool_disable"))
# 增强工具显示
if "enhanced_tool_display" in data:
base["enhanced_tool_display"] = bool(data.get("enhanced_tool_display"))
else:
base["enhanced_tool_display"] = bool(base.get("enhanced_tool_display", True))
# 简略消息显示full完整原始内容/ brief一行概要
compact_msg = data.get("compact_message_display", base.get("compact_message_display"))
if isinstance(compact_msg, str) and compact_msg.strip().lower() in ("full", "brief"):
base["compact_message_display"] = compact_msg.strip().lower()
else:
base["compact_message_display"] = "full"
# 堆叠块隐藏边线
if "stacked_hide_borders" in data:
base["stacked_hide_borders"] = bool(data.get("stacked_hide_borders"))
else:
base["stacked_hide_borders"] = bool(base.get("stacked_hide_borders", False))
# Git 状态栏显示开关
if "show_git_status_bar" in data:
base["show_git_status_bar"] = bool(data.get("show_git_status_bar"))
else:
base["show_git_status_bar"] = bool(base.get("show_git_status_bar", True))
# 使用自定义称呼
if "use_custom_names" in data:
base["use_custom_names"] = bool(data.get("use_custom_names"))
else:
base["use_custom_names"] = bool(base.get("use_custom_names"))
# AGENTS.md 自动注入开关
if "agents_md_auto_inject" in data:
base["agents_md_auto_inject"] = bool(data.get("agents_md_auto_inject"))
else:
base["agents_md_auto_inject"] = bool(base.get("agents_md_auto_inject", False))
# 允许根目录创建文件开关
if "allow_root_file_creation" in data:
base["allow_root_file_creation"] = bool(data.get("allow_root_file_creation"))
else:
base["allow_root_file_creation"] = bool(base.get("allow_root_file_creation", False))
# 默认隐藏工作区
if "default_hide_workspace" in data:
base["default_hide_workspace"] = bool(data.get("default_hide_workspace"))
else:
base["default_hide_workspace"] = bool(base.get("default_hide_workspace", False))
# 主题配色
theme_value = data.get("theme", base.get("theme"))
if isinstance(theme_value, str) and theme_value in ALLOWED_THEMES:
base["theme"] = theme_value
elif base.get("theme") not in ALLOWED_THEMES:
base["theme"] = "classic"
# 目标模式:审核模式
goal_review_mode = data.get("goal_review_mode", base.get("goal_review_mode"))
if isinstance(goal_review_mode, str) and goal_review_mode in ALLOWED_GOAL_REVIEW_MODES:
base["goal_review_mode"] = goal_review_mode
elif base.get("goal_review_mode") not in ALLOWED_GOAL_REVIEW_MODES:
base["goal_review_mode"] = "readonly"
# 目标模式:结束方式(多选)
if "goal_end_conditions" in data:
base["goal_end_conditions"] = _sanitize_goal_end_conditions(data.get("goal_end_conditions"))
else:
base["goal_end_conditions"] = _sanitize_goal_end_conditions(base.get("goal_end_conditions"))
# 目标模式:最大轮数
if "goal_max_turns" in data:
base["goal_max_turns"] = _sanitize_optional_int(
data.get("goal_max_turns"), min_value=GOAL_MAX_TURNS_MIN, max_value=GOAL_MAX_TURNS_MAX
) or GOAL_MAX_TURNS_DEFAULT
else:
base["goal_max_turns"] = _sanitize_optional_int(
base.get("goal_max_turns"), min_value=GOAL_MAX_TURNS_MIN, max_value=GOAL_MAX_TURNS_MAX
) or GOAL_MAX_TURNS_DEFAULT
# 目标模式:累计 token 上限(可空)
if "goal_max_tokens" in data:
base["goal_max_tokens"] = _sanitize_optional_int(
data.get("goal_max_tokens"), min_value=GOAL_MAX_TOKENS_MIN, max_value=GOAL_MAX_TOKENS_MAX
)
else:
base["goal_max_tokens"] = _sanitize_optional_int(
base.get("goal_max_tokens"), min_value=GOAL_MAX_TOKENS_MIN, max_value=GOAL_MAX_TOKENS_MAX
)
return base
def _sanitize_goal_end_conditions(value: Any) -> list:
"""清洗目标模式结束方式列表,保证至少包含 max_turns。"""
cleaned: list = []
if isinstance(value, list):
for item in value:
if isinstance(item, str) and item in ALLOWED_GOAL_END_CONDITIONS and item not in cleaned:
cleaned.append(item)
if "max_turns" not in cleaned:
cleaned.insert(0, "max_turns")
return cleaned
def _sanitize_skills(value: Any) -> Optional[list]:
"""Sanitize enabled skills list / 清洗启用技能列表。"""
if value is None:
return None
if not isinstance(value, list):
return []
cleaned: list = []
seen = set()
for item in value:
if not isinstance(item, str):
continue
skill_id = item.strip()
if not skill_id or skill_id in seen:
continue
cleaned.append(skill_id)
seen.add(skill_id)
return cleaned
def save_personalization_config(base_dir: PathLike, payload: Dict[str, Any]) -> Dict[str, Any]:
"""Persist sanitized personalization config and return it."""
existing = load_personalization_config(base_dir)
config = sanitize_personalization_payload(payload, fallback=existing)
validate_context_compression_settings(config)
path = _to_path(base_dir)
_ensure_parent(path)
with open(path, "w", encoding="utf-8") as f:
json.dump(config, f, ensure_ascii=False, indent=2)
return config
def _sanitize_optional_int(value: Any, *, min_value: int, max_value: int) -> Optional[int]:
if value is None or value == "":
return None
if isinstance(value, bool):
return None
try:
parsed = int(value)
except (TypeError, ValueError):
return None
if parsed < min_value:
return min_value
if parsed > max_value:
return max_value
return parsed
def resolve_context_compression_settings(config: Optional[Dict[str, Any]]) -> Dict[str, int]:
data = config or {}
shallow_trigger = _sanitize_optional_int(
data.get("shallow_compress_trigger_tokens"),
min_value=MIN_COMPRESSION_TRIGGER_TOKENS,
max_value=MAX_COMPRESSION_TRIGGER_TOKENS,
) or DEFAULT_SHALLOW_COMPRESS_TRIGGER_TOKENS
shallow_keep_recent = _sanitize_optional_int(
data.get("shallow_compress_keep_recent_tools"),
min_value=MIN_SHALLOW_KEEP_RECENT_TOOLS,
max_value=MAX_SHALLOW_KEEP_RECENT_TOOLS,
)
if shallow_keep_recent is None:
shallow_keep_recent = DEFAULT_SHALLOW_COMPRESS_KEEP_RECENT_TOOLS
shallow_keep_user_turn = _sanitize_optional_int(
data.get("shallow_compress_keep_user_turn_tools"),
min_value=MIN_SHALLOW_KEEP_USER_TURN_TOOLS,
max_value=MAX_SHALLOW_KEEP_USER_TURN_TOOLS,
)
if shallow_keep_user_turn is None:
shallow_keep_user_turn = DEFAULT_SHALLOW_KEEP_USER_TURN_TOOLS
shallow_max_replace = _sanitize_optional_int(
data.get("shallow_compress_max_replace_per_round"),
min_value=MIN_SHALLOW_MAX_REPLACE_PER_ROUND,
max_value=MAX_SHALLOW_MAX_REPLACE_PER_ROUND,
) or DEFAULT_SHALLOW_COMPRESS_MAX_REPLACE_PER_ROUND
shallow_trigger_interval = _sanitize_optional_int(
data.get("shallow_compress_trigger_tool_calls_interval"),
min_value=MIN_SHALLOW_TRIGGER_TOOL_CALLS_INTERVAL,
max_value=MAX_SHALLOW_TRIGGER_TOOL_CALLS_INTERVAL,
) or DEFAULT_SHALLOW_COMPRESS_TRIGGER_TOOL_CALLS_INTERVAL
deep_trigger = _sanitize_optional_int(
data.get("deep_compress_trigger_tokens"),
min_value=MIN_COMPRESSION_TRIGGER_TOKENS,
max_value=MAX_COMPRESSION_TRIGGER_TOKENS,
) or DEFAULT_DEEP_COMPRESS_TRIGGER_TOKENS
if deep_trigger <= shallow_trigger:
deep_trigger = shallow_trigger + 1
return {
"shallow_trigger_tokens": shallow_trigger,
"shallow_keep_recent_tools": shallow_keep_recent,
"shallow_keep_user_turn_tools": shallow_keep_user_turn,
"shallow_max_replace_per_round": shallow_max_replace,
"shallow_trigger_tool_calls_interval": shallow_trigger_interval,
"deep_trigger_tokens": deep_trigger,
}
def validate_context_compression_settings(config: Optional[Dict[str, Any]]) -> None:
data = config or {}
shallow_trigger = _sanitize_optional_int(
data.get("shallow_compress_trigger_tokens"),
min_value=MIN_COMPRESSION_TRIGGER_TOKENS,
max_value=MAX_COMPRESSION_TRIGGER_TOKENS,
) or DEFAULT_SHALLOW_COMPRESS_TRIGGER_TOKENS
deep_trigger = _sanitize_optional_int(
data.get("deep_compress_trigger_tokens"),
min_value=MIN_COMPRESSION_TRIGGER_TOKENS,
max_value=MAX_COMPRESSION_TRIGGER_TOKENS,
) or DEFAULT_DEEP_COMPRESS_TRIGGER_TOKENS
if deep_trigger <= shallow_trigger:
raise ValueError("深压缩触发上下文必须大于浅压缩触发上下文")
def _load_human_like_prompt() -> str:
"""加载拟人化风格提示词文件。"""
prompt = _load_prompt_file("human_like_style.txt")
if prompt:
return prompt
# 默认提示词
return (
"不要「像AI一样说话」而是尽可能的模仿真人的说话方式或是模仿用户输入的说话方式。\n"
"尽可能不要用markdown格式尤其是\n"
"- ###、##、#等一二三级标题\n"
"- 等列表形式\n"
"- 不必要的** **加粗\n"
"- 刻意的分点分层次的回答\n"
"用户如果不输入emoji就不要使用emoji回答。"
)
def _load_auto_style_prompt() -> str:
"""加载自动交流风格提示词文件。"""
return _load_prompt_file("auto_style.txt")
def _load_prompt_file(filename: str) -> str:
from config.paths import PROMPTS_DIR
try:
prompt_path = Path(PROMPTS_DIR) / filename
if prompt_path.exists():
return prompt_path.read_text(encoding="utf-8").strip()
except Exception:
pass
return ""
def build_personalization_prompt(
config: Optional[Dict[str, Any]],
include_header: bool = True
) -> Optional[str]:
"""Generate the personalization prompt text based on config."""
if not config or not config.get("enabled"):
return None
lines = []
if include_header:
lines.append("用户的个性化数据,请回答时务必参照这些信息")
if config.get("self_identify"):
lines.append(f"用户希望你自称:{config['self_identify']}")
if config.get("user_name"):
lines.append(f"用户希望你称呼为:{config['user_name']}")
if config.get("profession"):
lines.append(f"用户的职业是:{config['profession']}")
if config.get("tone"):
lines.append(f"用户希望你使用 {config['tone']} 的语气与TA交流")
considerations: Iterable[str] = config.get("considerations") or []
considerations = [item for item in considerations if item]
if considerations:
lines.append("用户希望你在回答问题时必须考虑的信息是:")
for idx, item in enumerate(considerations, 1):
lines.append(f"{idx}. {item}")
conversation_continuity = str(config.get("conversation_continuity") or "medium").strip().lower()
if conversation_continuity == "high":
lines.append("\n【对话连续性:高】")
lines.append(
"当前对话应积极延续用户长期背景。回答时优先参考用户过往对话、长期记忆和项目背景;"
"遇到可能相关的问题,应主动搜索/回顾历史对话;发现值得长期保留的偏好、项目事实、"
"稳定经验或用户明确要求记住的内容,应主动调用记忆工具记录。"
)
elif conversation_continuity == "low":
lines.append("\n【对话连续性:低】")
lines.append(
"当前对话应尽量独立。不要主动回顾之前的对话,不要主动调用记忆或历史对话工具,"
"不要主动参考用户历史记忆来回答;只有用户明确要求、当前任务必须依赖历史,"
"或安全/连续性需要时,才可搜索历史或读取记忆。"
)
else:
lines.append("\n【对话连续性:中】")
lines.append(
"当前对话优先,但可以在适当时候参考历史。若用户问题明显与过往偏好、项目背景或未完成事项有关,"
"可以调用记忆或历史对话工具;不要为了普通问题主动翻历史。"
)
communication_style = str(config.get("communication_style") or "default").strip().lower()
if communication_style == "auto":
auto_style_prompt = _load_auto_style_prompt()
if auto_style_prompt:
lines.append("\n【交流风格:自动】")
lines.append(auto_style_prompt)
elif communication_style == "human_like":
human_like_prompt = _load_human_like_prompt()
if human_like_prompt:
lines.append("\n【交流风格要求】")
lines.append(human_like_prompt)
if len(lines) == (1 if include_header else 0):
# 没有任何有效内容时不注入
return None
return "\n".join(lines)
def _sanitize_short_field(value: Optional[str]) -> str:
if not value:
return ""
text = str(value).strip()
if not text:
return ""
return text[:MAX_SHORT_FIELD_LENGTH]
def _sanitize_considerations(value: Any) -> list:
if not isinstance(value, list):
return []
cleaned = []
for item in value:
if not isinstance(item, str):
continue
text = item.strip()
if not text:
continue
cleaned.append(text[:MAX_CONSIDERATION_LENGTH])
if len(cleaned) >= MAX_CONSIDERATION_ITEMS:
break
return cleaned
def _sanitize_thinking_interval(value: Any) -> Optional[int]:
if value is None or value == "":
return None
try:
interval = int(value)
except (TypeError, ValueError):
return None
interval = max(THINKING_INTERVAL_MIN, min(THINKING_INTERVAL_MAX, interval))
if interval == THINKING_FAST_INTERVAL:
return None
return interval
def _sanitize_tool_categories(value: Any, allowed: set) -> list:
if not isinstance(value, list):
return []
result = []
dynamic_prefixes = ("mcp_server__",)
for item in value:
if not isinstance(item, str):
continue
candidate = item.strip()
if not candidate:
continue
dynamic_ok = candidate == "custom" or any(candidate.startswith(prefix) for prefix in dynamic_prefixes)
if candidate not in allowed and not dynamic_ok:
continue
if candidate not in result:
result.append(candidate)
return result
def _sanitize_run_mode(value: Any) -> Optional[str]:
if value is None:
return None
if isinstance(value, str):
candidate = value.strip().lower()
if candidate in ALLOWED_RUN_MODES:
return candidate
return None