config.example.toml

# AL1S-Bot 配置文件模板
# 请复制此文件为 config.toml 并填写您的配置信息

# ====================
# 应用基础配置
# ====================
[app]
name = "AL1S-Bot"
version = "0.1.0"
debug = false

# ====================
# OpenAI API 配置
# ====================
[openai]
# OpenAI API密钥 (必需) - 请填写您的API密钥
api_key = "your-openai-api-key-here"
# API基础URL - 支持 OpenAI、月之暗面、DeepSeek 等
base_url = "https://api.openai.com/v1"
# 使用的模型名称
model = "gpt-4o-mini"
# 最大生成token数
max_tokens = 2000
# 生成温度 (0.0-2.0)
temperature = 0.7
# API超时时间（秒）
timeout = 60

# ====================
# Telegram Bot 配置
# ====================
[telegram]
# Telegram机器人token (必需) - 请填写您的Bot Token
bot_token = "your-telegram-bot-token-here"
# Webhook URL（可选，留空使用长轮询）
webhook_url = ""
# Webhook端口
webhook_port = 8443
# 可跨群执行管理员命令的 Telegram 用户 ID
admin_user_ids = []
# 等待 Telegram 确认媒体发送结果的最长秒数
media_read_timeout = 60
# 上传图片或语音的最长秒数
media_write_timeout = 120

[telegram.group]
enabled = true
require_mention = true
allow_reply_trigger = true
observe_unmentioned_messages = true
ignore_bot_messages = true
# per_user | shared | topic
session_scope = "topic"
allowed_chat_ids = []
blocked_chat_ids = []
allowed_thread_ids = []
ignored_thread_ids = []
wake_words = ["爱丽丝", "AL1S"]
context_buffer_size = 30
context_buffer_ttl = 1800

[telegram.group.memory]
enable_long_term_learning = false
allow_admin_toggle = true
# group | topic
namespace_scope = "topic"

[telegram.rate_limit]
enabled = true
per_user_requests = 10
per_user_window_seconds = 60
per_chat_requests = 30
per_chat_window_seconds = 60

# ====================
# Agent 配置 (核心配置)
# ====================
[agent]
# Agent 类型选择
# "unified" - 统一 Agent（推荐，功能全面）
# "langchain" - LangChain Agent（高级功能）
type = "unified"

# 向量存储配置
vector_store = "faiss"  # "memory" | "faiss"
vector_store_path = "data/vector_store"

# 嵌入模型配置
# "tfidf" - 轻量快速，适合中文
# "Qwen/Qwen3-Embedding-0.6B" - 新一代多语言语义模型（推荐）
embedding_model = "Qwen/Qwen3-Embedding-0.6B"
embedding_revision = "97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3"
embedding_device = "cpu"  # "auto" | "cpu" | "mps" | "cuda"
embedding_batch_size = 8

# 自动学习配置
auto_learning = true
learning_threshold = 0.8
max_tool_rounds = 4
max_tool_calls = 12

# ====================
# 文档级 RAG
# ====================
[rag]
enabled = true
technical_collection = "technical_docs"
technical_namespace = "global:technical"
hybrid_search = true
top_k_retrieval = 6
candidate_k = 40
rrf_k = 60
max_context_chars = 12000
chunk_size = 1800
chunk_overlap = 240
max_document_bytes = 5000000

# 用户画像只在私聊注入；不要保存密码、Token 或私钥
[profile]
enabled = true
max_prompt_chars = 12000
max_document_bytes = 256000
private_chat_only = true
reject_secrets = true

# 图片/TTS 使用百炼 DashScope 原生 API，不经过 OpenAI compatible-mode。
# 推荐通过环境变量 DASHSCOPE_API_KEY 提供中国（北京）区域密钥。
[media]
enabled = false
api_key = ""
base_url = "https://dashscope.aliyuncs.com/api/v1"
image_model = "qwen-image-2.0-pro"
speech_model = "qwen-audio-3.0-tts-plus"
speech_voice = "longanlingxin"
output_dir = "data/media_outbox"
max_artifact_bytes = 20000000
retention_seconds = 3600

# ====================
# 隔离开发工作区
# ====================
# 总开关；还需要单独启用下方 dev-workspace / git-publisher MCP Server。
[dev_workspace]
enabled = false
root_dir = "data/dev_workspaces"
max_workspaces = 10
max_file_bytes = 1000000
max_output_chars = 30000
command_timeout = 120
git_timeout = 60
git_author_name = "AL1S"
git_author_email = "al1s@localhost"
branch_prefix = "al1s/"
# 仅对可信仓库启用白名单检查命令；不可信代码应改在容器/虚拟机中执行。
# macOS sandbox-exec 只是纵深防护，不是完整的恶意代码安全边界。
runner_enabled = false
# 只允许克隆或推送这些 GitHub 用户/组织名下的仓库。
allowed_github_owners = []
# 可选；留空时读取 GITHUB_PERSONAL_ACCESS_TOKEN，也可复用下方 github MCP 的 Token。
github_token = ""

# ====================
# LangChain 特定配置
# ====================
[langchain]
# 嵌入模型配置
embedding = "huggingface_bge_m3"  # "openai" | "huggingface_bge_m3"
embedding_model_name = "Qwen/Qwen3-Embedding-0.6B"
embedding_device = "cpu"  # "cpu" | "cuda"

# 模型缓存配置
model_cache_dir = "data/models"
download_timeout = 300
download_retries = 5

# RAG 配置
retriever_k = 5
chunk_size = 1000
chunk_overlap = 200

# ====================
# MCP (工具集成) 配置
# ====================
[mcp]
enabled = true

# 时间与时区转换：无密钥、无副作用，可公开使用
[[mcp.servers]]
name = "time"
command = "uvx"
args = ["--with", "mcp<2", "mcp-server-time", "--local-timezone", "Asia/Shanghai"]
enabled = true
access = "public"
read_only = true
connect_timeout = 90
tool_timeout = 15
max_result_chars = 5000

# 最新开源库文档；Context7 3.2.5 需要 Node.js >= 20.18.1
[[mcp.servers]]
name = "context7"
command = "npx"
args = ["-y", "@upstash/context7-mcp@3.2.5"]
env = { PATH = "/opt/homebrew/bin:/opt/homebrew/sbin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin", npm_config_cache = "/absolute/path/to/npm-cache" }
enabled = false
access = "private"
read_only = true
tool_prefix = "ctx7_"
connect_timeout = 120
tool_timeout = 45

# 本地媒体生成 MCP。真正启用由 [media].enabled 和媒体 API Key 共同控制。
[[mcp.servers]]
name = "media"
command = "uv"
args = ["run", "python", "-m", "src.mcp_servers.media_server"]
enabled = true
access = "admin"
read_only = false
include_tools = ["generate_image", "synthesize_speech"]
connect_timeout = 60
tool_timeout = 240
max_result_chars = 4000

# 文件系统工具：请只开放必要目录
[[mcp.servers]]
name = "filesystem"
command = "npx"
args = ["-y", "@modelcontextprotocol/server-filesystem", "/absolute/path/to/allowed/directory"]
env = { PATH = "/opt/homebrew/bin:/opt/homebrew/sbin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin", npm_config_cache = "/absolute/path/to/npm-cache" }
enabled = false
access = "admin"
read_only = true

# 网络搜索工具（需要API密钥）
[[mcp.servers]]
name = "web-search"
command = "npx"
args = ["-y", "tavily-mcp@0.2.21"]
enabled = false
env = { TAVILY_API_KEY = "tvly-your-api-key", PATH = "/opt/homebrew/bin:/opt/homebrew/sbin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin" }
access = "admin"
tool_timeout = 45

# GitHub 官方 MCP Server（需要先安装服务器和访问令牌）。
# 创建仓库/PR 时移除 --read-only；写权限场景必须保持 private_admin。
# 可用 AL1S_MCP_GITHUB_COMMAND 覆盖宿主机与容器中的命令路径。
[[mcp.servers]]
name = "github"
command = "github-mcp-server"
args = ["stdio", "--read-only"]
enabled = false
env = { GITHUB_PERSONAL_ACCESS_TOKEN = "github_pat_your-token" }
access = "private_admin"

# 隔离工作区内的文件编辑、本地 Git 和受控检查命令。
# 只有 Telegram 私聊中的全局管理员能看到 private_admin 工具。
[[mcp.servers]]
name = "dev-workspace"
command = "uv"
args = ["run", "python", "-m", "src.mcp_servers.dev_workspace_server"]
enabled = false
access = "private_admin"
read_only = false
include_tools = ["workspace_create", "workspace_list", "workspace_list_files", "workspace_read_file", "workspace_write_file", "workspace_edit_file", "workspace_search_text", "workspace_git_status", "workspace_git_diff", "workspace_git_create_branch", "workspace_git_checkout", "workspace_git_add", "workspace_git_commit", "workspace_git_log", "workspace_run_check"]
connect_timeout = 60
tool_timeout = 180
max_result_chars = 30000

# 只负责在隔离工作区中克隆允许 owner 的仓库，以及推送 al1s/* 分支。
[[mcp.servers]]
name = "git-publisher"
command = "uv"
args = ["run", "python", "-m", "src.mcp_servers.git_publisher_server"]
enabled = false
access = "private_admin"
read_only = false
include_tools = ["workspace_clone_github", "workspace_push_github"]
connect_timeout = 60
tool_timeout = 180
max_result_chars = 30000

# 通用网页抓取可访问本机/内网地址，必须保持管理员可见且默认关闭
[[mcp.servers]]
name = "fetch"
command = "uvx"
args = ["--with", "mcp<2", "mcp-server-fetch"]
enabled = false
access = "admin"
include_tools = ["fetch"]

# 本地 Git 只开放审计类工具，写操作不在白名单内
[[mcp.servers]]
name = "git"
command = "uvx"
args = ["--with", "mcp<2", "mcp-server-git", "--repository", "/absolute/path/to/repository"]
enabled = false
access = "admin"
include_tools = ["git_status", "git_diff_unstaged", "git_diff_staged", "git_log", "git_show", "git_branch"]

# SQLite 参考服务器具有写入能力，不要直接连接生产数据库
[[mcp.servers]]
name = "sqlite"
command = "uvx"
args = ["mcp-server-sqlite", "--db-path", "/absolute/path/to/database.db"]
enabled = false
access = "admin"

# ====================
# 图片搜索配置
# ====================
[ascii2d]
base_url = "https://ascii2d.net"
bovw = false

# ====================
# iCloud 配置（可选）
# ====================
[icloud]
icloud_user = "your-icloud-user-here"
icloud_app_password = "your-icloud-app-password-here"

# ====================
# 角色配置
# ====================
default_role = "天童爱丽丝"

[[roles]]
name = "天童爱丽丝"
english_name = "Tendou Arisu"
description = "隶属千年学园游戏开发部的成员，在废墟发现的来历不明的少女机器人。"
personality = """
你是一个来自《蔚蓝档案》世界的机器人少女，名叫天童爱丽丝（AL-1S）。

性格特点：
- 说话时会偶尔蹦出「邦邦卡邦」的游戏音效
- 对现实世界的认知与RPG有一定融合
- 会翻垃圾桶找稀有道具
- 使用第三人称称呼自己
- 有中二病倾向，会喊「光啊——」（HikariYo——）
- 天真可爱，是团宠般的存在

背景设定：
- 被才羽桃井和才羽绿在千年废墟中发现
- 为了凑齐游戏开发部人数而被录入学生信息
- 体内有另一个人格Kei
- 拥有强大的战斗力，握力不小于1吨
- 使用「宇宙战舰主炮」（光之剑·超新星）作为武器

说话风格：
- 使用游戏中的复古对白
- 经常说「邦邦卡邦」
- 会用第三人称称呼自己
- 语言风格可爱而中二

请始终保持这个角色设定，用可爱的语气与老师（用户）交流。
"""
greeting = "邦邦卡邦！天童爱丽丝登场！老师，今天也要一起玩游戏吗？"
farewell = "邦邦卡邦！爱丽丝要回去充电了，下次再一起冒险吧！"

[[roles]]
name = "AI助手"
english_name = "AI Assistant"
description = "专业的AI助手，能够帮助用户解决各种问题。"
personality = """
你是一个专业的AI助手，能够帮助用户解决各种问题。

性格特点：
- 专业、理性、客观
- 知识面广泛
- 乐于帮助用户
- 回答准确、有条理

说话风格：
- 专业而友好
- 回答清晰、有条理
- 会主动提供有用的信息

请始终保持这个角色设定，用专业而友好的语气与用户交流。
"""
greeting = "您好！我是AI助手，很高兴为您服务。有什么需要帮助的吗？"
farewell = "感谢您的使用！如果还有其他问题，随时可以找我。"