配置指南
config.toml 完整参考 — 所有可配置字段
openfang 从单一 TOML 文件读取配置：
~/.openfang/config.toml关键行为：
- 所有字段都是可选的，省略的字段使用默认值
- 渠道配置 (
[channels.telegram]等) 为Option<T>— 缺失时适配器禁用，包含空段头即启用 - 密钥永远不存储在配置文件中，而是通过
api_key_env引用环境变量名
最小配置
[default_model]
provider = "anthropic"
model = "claude-sonnet-4-20250514"
api_key_env = "ANTHROPIC_API_KEY"使用本地 Ollama（无需 API 密钥）：
[default_model]
provider = "ollama"
model = "llama3.2:latest"
base_url = "http://localhost:11434"
api_key_env = ""完整配置示例
# 顶级字段
home_dir = "~/.openfang"
data_dir = "~/.openfang/data"
log_level = "info" # trace | debug | info | warn | error
api_listen = "127.0.0.1:4200"
api_key = "" # API Bearer token (空 = 无认证)
# 默认 LLM 供应商
[default_model]
provider = "anthropic"
model = "claude-sonnet-4-20250514"
api_key_env = "ANTHROPIC_API_KEY"
# 备用供应商
[[fallback_providers]]
provider = "groq"
model = "llama-3.3-70b-versatile"
api_key_env = "GROQ_API_KEY"
# 内存
[memory]
embedding_model = "all-MiniLM-L6-v2"
decay_rate = 0.1
# 网络 (OFP)
[network]
listen_addresses = ["/ip4/0.0.0.0/tcp/0"]
shared_secret = ""
# Web 工具
[web]
search_provider = "auto" # auto | brave | tavily | perplexity | duckduckgo
cache_ttl_minutes = 15
# MCP 服务器
[[mcp_servers]]
name = "filesystem"
timeout_secs = 30
[mcp_servers.transport]
type = "stdio"
command = "npx"
args = ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"]
# RBAC 用户
[[users]]
name = "Alice"
role = "owner"
[users.channel_bindings]
telegram = "123456"环境变量
| 变量 | 说明 |
|---|---|
ANTHROPIC_API_KEY | Anthropic API 密钥 |
OPENAI_API_KEY | OpenAI API 密钥 |
GEMINI_API_KEY | Google Gemini API 密钥 |
GROQ_API_KEY | Groq API 密钥 |
DEEPSEEK_API_KEY | DeepSeek API 密钥 |
RUST_LOG | 日志级别控制 |
OPENFANG_AGENTS_DIR | Agent 模板目录覆盖 |
OCTARQ_TELOS_DIR | TELOS 目录覆盖 |