| 1 | """Configuration schema using Pydantic.""" |
| 2 | |
| 3 | from pathlib import Path |
| 4 | from typing import Literal |
| 5 | |
| 6 | from pydantic import AliasChoices, BaseModel, ConfigDict, Field |
| 7 | from pydantic.alias_generators import to_camel |
| 8 | from pydantic_settings import BaseSettings |
| 9 | |
| 10 | from nanobot.cron.types import CronSchedule |
| 11 | |
| 12 | |
| 13 | class Base(BaseModel): |
| 14 | """Base model that accepts both camelCase and snake_case keys.""" |
| 15 | |
| 16 | model_config = ConfigDict(alias_generator=to_camel, populate_by_name=True) |
| 17 | |
| 18 | class ChannelsConfig(Base): |
| 19 | """Configuration for chat channels. |
| 20 | |
| 21 | Built-in and plugin channel configs are stored as extra fields (dicts). |
| 22 | Each channel parses its own config in __init__. |
| 23 | Per-channel "streaming": true enables streaming output (requires send_delta impl). |
| 24 | """ |
| 25 | |
| 26 | model_config = ConfigDict(extra="allow") |
| 27 | |
| 28 | send_progress: bool = True # stream agent's text progress to the channel |
| 29 | send_tool_hints: bool = False # stream tool-call hints (e.g. read_file("…")) |
| 30 | send_max_retries: int = Field(default=3, ge=0, le=10) # Max delivery attempts (initial send included) |
| 31 | transcription_provider: str = "groq" # Voice transcription backend: "groq" or "openai" |
| 32 | transcription_language: str | None = Field(default=None, pattern=r"^[a-z]{2,3}$") # Optional ISO-639-1 hint for audio transcription |
| 33 | |
| 34 | |
| 35 | class DreamConfig(Base): |
| 36 | """Dream memory consolidation configuration.""" |
| 37 | |
| 38 | _HOUR_MS = 3_600_000 |
| 39 | |
| 40 | interval_h: int = Field(default=2, ge=1) # Every 2 hours by default |
| 41 | cron: str | None = Field(default=None, exclude=True) # Legacy compatibility override |
| 42 | model_override: str | None = Field( |
| 43 | default=None, |
| 44 | validation_alias=AliasChoices("modelOverride", "model", "model_override"), |
| 45 | ) # Optional Dream-specific model override |
| 46 | max_batch_size: int = Field(default=20, ge=1) # Max history entries per run |
| 47 | # Bumped from 10 to 15 in #3212 (exp002: +30% dedup, no accuracy loss; >15 plateaus). |
| 48 | max_iterations: int = Field(default=15, ge=1) # Max tool calls per Phase 2 |
| 49 | # Per-line git-blame age annotation in Phase 1 prompt (see #3212). Default |
| 50 | # on — set to False to feed MEMORY.md raw if a specific LLM reacts poorly |
| 51 | # to the `← Nd` suffix or you want deterministic, git-independent prompts. |
| 52 | annotate_line_ages: bool = True |
| 53 | |
| 54 | def build_schedule(self, timezone: str) -> CronSchedule: |
| 55 | """Build the runtime schedule, preferring the legacy cron override if present.""" |
| 56 | if self.cron: |
| 57 | return CronSchedule(kind="cron", expr=self.cron, tz=timezone) |
| 58 | return CronSchedule(kind="every", every_ms=self.interval_h * self._HOUR_MS) |
| 59 | |
| 60 | def describe_schedule(self) -> str: |
| 61 | """Return a human-readable summary for logs and startup output.""" |
| 62 | if self.cron: |
| 63 | return f"cron {self.cron} (legacy)" |
| 64 | hours = self.interval_h |
| 65 | return f"every {hours}h" |
| 66 | |
| 67 | |
| 68 | class AgentDefaults(Base): |
| 69 | """Default agent configuration.""" |
| 70 | |
| 71 | workspace: str = "~/.nanobot/workspace" |
| 72 | model: str = "anthropic/claude-opus-4-5" |
| 73 | provider: str = ( |
| 74 | "auto" # Provider name (e.g. "anthropic", "openrouter") or "auto" for auto-detection |
| 75 | ) |
| 76 | max_tokens: int = 8192 |
| 77 | context_window_tokens: int = 65_536 |
| 78 | context_block_limit: int | None = None |
| 79 | temperature: float = 0.1 |
| 80 | max_tool_iterations: int = 200 |
| 81 | max_tool_result_chars: int = 16_000 |
| 82 | max_concurrent_requests: int = Field(default=3, ge=0) |
| 83 | provider_retry_mode: Literal["standard", "persistent"] = "standard" |
| 84 | reasoning_effort: str | None = None # low / medium / high / adaptive - enables LLM thinking mode |
| 85 | timezone: str = "UTC" # IANA timezone, e.g. "Asia/Shanghai", "America/New_York" |
| 86 | unified_session: bool = False # Share one session across all channels (single-user multi-device) |
| 87 | disabled_skills: list[str] = Field(default_factory=list) # Skill names to exclude from loading (e.g. ["summarize", "skill-creator"]) |
| 88 | session_ttl_minutes: int = Field( |
| 89 | default=0, |
| 90 | ge=0, |
| 91 | validation_alias=AliasChoices("idleCompactAfterMinutes", "sessionTtlMinutes"), |
| 92 | serialization_alias="idleCompactAfterMinutes", |
| 93 | ) # Auto-compact idle threshold in minutes (0 = disabled) |
| 94 | dream: DreamConfig = Field(default_factory=DreamConfig) |
| 95 | |
| 96 | |
| 97 | class AgentsConfig(Base): |
| 98 | """Agent configuration.""" |
| 99 | |
| 100 | defaults: AgentDefaults = Field(default_factory=AgentDefaults) |
| 101 | |
| 102 | |
| 103 | class ProviderConfig(Base): |
| 104 | """LLM provider configuration.""" |
| 105 | |
| 106 | api_key: str | None = None |
| 107 | api_base: str | None = None |
| 108 | extra_headers: dict[str, str] | None = None # Custom headers (e.g. APP-Code for AiHubMix) |
| 109 | |
| 110 | |
| 111 | class ProvidersConfig(Base): |
| 112 | """Configuration for LLM providers.""" |
| 113 | |
| 114 | custom: ProviderConfig = Field(default_factory=ProviderConfig) # Any OpenAI-compatible endpoint |
| 115 | azure_openai: ProviderConfig = Field(default_factory=ProviderConfig) # Azure OpenAI (model = deployment name) |
| 116 | anthropic: ProviderConfig = Field(default_factory=ProviderConfig) |
| 117 | openai: ProviderConfig = Field(default_factory=ProviderConfig) |
| 118 | openrouter: ProviderConfig = Field(default_factory=ProviderConfig) |
| 119 | deepseek: ProviderConfig = Field(default_factory=ProviderConfig) |
| 120 | groq: ProviderConfig = Field(default_factory=ProviderConfig) |
| 121 | zhipu: ProviderConfig = Field(default_factory=ProviderConfig) |
| 122 | dashscope: ProviderConfig = Field(default_factory=ProviderConfig) |
| 123 | vllm: ProviderConfig = Field(default_factory=ProviderConfig) |
| 124 | ollama: ProviderConfig = Field(default_factory=ProviderConfig) # Ollama local models |
| 125 | lm_studio: ProviderConfig = Field(default_factory=ProviderConfig) # LM Studio local models |
| 126 | ovms: ProviderConfig = Field(default_factory=ProviderConfig) # OpenVINO Model Server (OVMS) |
| 127 | gemini: ProviderConfig = Field(default_factory=ProviderConfig) |
| 128 | moonshot: ProviderConfig = Field(default_factory=ProviderConfig) |
| 129 | minimax: ProviderConfig = Field(default_factory=ProviderConfig) |
| 130 | minimax_anthropic: ProviderConfig = Field(default_factory=ProviderConfig) # MiniMax Anthropic endpoint (thinking) |
| 131 | mistral: ProviderConfig = Field(default_factory=ProviderConfig) |
| 132 | stepfun: ProviderConfig = Field(default_factory=ProviderConfig) # Step Fun (阶跃星辰) |
| 133 | xiaomi_mimo: ProviderConfig = Field(default_factory=ProviderConfig) # Xiaomi MIMO (小米) |
| 134 | aihubmix: ProviderConfig = Field(default_factory=ProviderConfig) # AiHubMix API gateway |
| 135 | siliconflow: ProviderConfig = Field(default_factory=ProviderConfig) # SiliconFlow (硅基流动) |
| 136 | volcengine: ProviderConfig = Field(default_factory=ProviderConfig) # VolcEngine (火山引擎) |
| 137 | volcengine_coding_plan: ProviderConfig = Field(default_factory=ProviderConfig) # VolcEngine Coding Plan |
| 138 | byteplus: ProviderConfig = Field(default_factory=ProviderConfig) # BytePlus (VolcEngine international) |
| 139 | byteplus_coding_plan: ProviderConfig = Field(default_factory=ProviderConfig) # BytePlus Coding Plan |
| 140 | openai_codex: ProviderConfig = Field(default_factory=ProviderConfig, exclude=True) # OpenAI Codex (OAuth) |
| 141 | github_copilot: ProviderConfig = Field(default_factory=ProviderConfig, exclude=True) # Github Copilot (OAuth) |
| 142 | qianfan: ProviderConfig = Field(default_factory=ProviderConfig) # Qianfan (百度千帆) |
| 143 | |
| 144 | |
| 145 | class HeartbeatConfig(Base): |
| 146 | """Heartbeat service configuration.""" |
| 147 | |
| 148 | enabled: bool = True |
| 149 | interval_s: int = 30 * 60 # 30 minutes |
| 150 | keep_recent_messages: int = 8 |
| 151 | |
| 152 | |
| 153 | class ApiConfig(Base): |
| 154 | """OpenAI-compatible API server configuration.""" |
| 155 | |
| 156 | host: str = "127.0.0.1" # Safer default: local-only bind. |
| 157 | port: int = 8900 |
| 158 | timeout: float = 120.0 # Per-request timeout in seconds. |
| 159 | public_base_url: str | None = None # Reserved for external OpenAI-compatible API deployments. |
| 160 | |
| 161 | |
| 162 | class GatewayConfig(Base): |
| 163 | """Gateway/server configuration.""" |
| 164 | |
| 165 | host: str = "127.0.0.1" # Safer default: local-only bind. |
| 166 | port: int = 18790 |
| 167 | heartbeat: HeartbeatConfig = Field(default_factory=HeartbeatConfig) |
| 168 | |
| 169 | |
| 170 | class WebSearchConfig(Base): |
| 171 | """Web search tool configuration.""" |
| 172 | |
| 173 | provider: str = "duckduckgo" # brave, tavily, duckduckgo, searxng, jina, kagi |
| 174 | api_key: str = "" |
| 175 | base_url: str = "" # SearXNG base URL |
| 176 | max_results: int = 5 |
| 177 | timeout: int = 30 # Wall-clock timeout (seconds) for search operations |
| 178 | |
| 179 | |
| 180 | class WebToolsConfig(Base): |
| 181 | """Web tools configuration.""" |
| 182 | |
| 183 | enable: bool = True |
| 184 | proxy: str | None = ( |
| 185 | None # HTTP/SOCKS5 proxy URL, e.g. "http://127.0.0.1:7890" or "socks5://127.0.0.1:1080" |
| 186 | ) |
| 187 | search: WebSearchConfig = Field(default_factory=WebSearchConfig) |
| 188 | |
| 189 | |
| 190 | class ExecToolConfig(Base): |
| 191 | """Shell exec tool configuration.""" |
| 192 | |
| 193 | enable: bool = True |
| 194 | timeout: int = 60 |
| 195 | path_append: str = "" |
| 196 | sandbox: str = "" # sandbox backend: "" (none) or "bwrap" |
| 197 | allowed_env_keys: list[str] = Field(default_factory=list) # Env var names to pass through to subprocess (e.g. ["GOPATH", "JAVA_HOME"]) |
| 198 | |
| 199 | class MCPServerConfig(Base): |
| 200 | """MCP server connection configuration (stdio or HTTP).""" |
| 201 | |
| 202 | type: Literal["stdio", "sse", "streamableHttp"] | None = None # auto-detected if omitted |
| 203 | command: str = "" # Stdio: command to run (e.g. "npx") |
| 204 | args: list[str] = Field(default_factory=list) # Stdio: command arguments |
| 205 | env: dict[str, str] = Field(default_factory=dict) # Stdio: extra env vars |
| 206 | url: str = "" # HTTP/SSE: endpoint URL |
| 207 | headers: dict[str, str] = Field(default_factory=dict) # HTTP/SSE: custom headers |
| 208 | tool_timeout: int = 30 # seconds before a tool call is cancelled |
| 209 | enabled_tools: list[str] = Field(default_factory=lambda: ["*"]) # Only register these tools; accepts raw MCP names or wrapped mcp_<server>_<tool> names; ["*"] = all tools; [] = no tools |
| 210 | |
| 211 | class MyToolConfig(Base): |
| 212 | """Self-inspection tool configuration.""" |
| 213 | |
| 214 | enable: bool = True # register the `my` tool (agent runtime state inspection) |
| 215 | allow_set: bool = False # let `my` modify loop state (read-only if False) |
| 216 | |
| 217 | |
| 218 | class EchoGeneratorConfig(Base): |
| 219 | """Echo generator backend configuration.""" |
| 220 | |
| 221 | base_url: str = "" # Local Echo Server origin (for example http://127.0.0.1:8221) |
| 222 | callback_base_url: str = "" # Local Agent callback origin (for example http://127.0.0.1:18791) |
| 223 | http_timeout_sec: float = 30.0 |
| 224 | |
| 225 | |
| 226 | class LocalFileStorageConfig(Base): |
| 227 | """Local filesystem storage exposed through the gateway.""" |
| 228 | |
| 229 | directory: str = "director/assets" |
| 230 | base_url: str = "" |
| 231 | route_prefix: str = "/api/assets" |
| 232 | |
| 233 | |
| 234 | class S3FileStorageConfig(Base): |
| 235 | """S3-compatible mapping for files sent to external services.""" |
| 236 | |
| 237 | endpoint_url: str = "" |
| 238 | public_base_url: str = "" |
| 239 | bucket: str = "" |
| 240 | region: str = "" |
| 241 | key_prefix: str = "" |
| 242 | addressing_style: Literal["auto", "virtual", "path"] = "auto" |
| 243 | access_key_id: str = "" |
| 244 | secret_access_key: str = "" |
| 245 | session_token: str = "" |
| 246 | |
| 247 | |
| 248 | class OutboundFileStorageConfig(Base): |
| 249 | """How locally managed files are exposed to external services.""" |
| 250 | |
| 251 | backend: Literal["inline", "s3"] = "inline" |
| 252 | s3: S3FileStorageConfig = Field(default_factory=S3FileStorageConfig) |
| 253 | |
| 254 | |
| 255 | class FileStorageConfig(Base): |
| 256 | """Local-first file storage selected independently from business logic.""" |
| 257 | |
| 258 | local: LocalFileStorageConfig = Field(default_factory=LocalFileStorageConfig) |
| 259 | outbound: OutboundFileStorageConfig = Field( |
| 260 | default_factory=OutboundFileStorageConfig |
| 261 | ) |
| 262 | |
| 263 | |
| 264 | class MemoryReviewConfig(Base): |
| 265 | """Local R2V Memory selection settings.""" |
| 266 | |
| 267 | enabled: bool = True |
| 268 | auto_approve: bool = False |
| 269 | candidate_count: int = Field(default=24, ge=4, le=24) |
| 270 | provider: str = "" |
| 271 | model: str = "" |
| 272 | |
| 273 | |
| 274 | class ToolsConfig(Base): |
| 275 | """Tools configuration.""" |
| 276 | |
| 277 | web: WebToolsConfig = Field(default_factory=WebToolsConfig) |
| 278 | exec: ExecToolConfig = Field(default_factory=ExecToolConfig) |
| 279 | my: MyToolConfig = Field(default_factory=MyToolConfig) |
| 280 | echo_generator: EchoGeneratorConfig = Field( |
| 281 | default_factory=EchoGeneratorConfig, |
| 282 | validation_alias=AliasChoices("echoGenerator", "directorRemote", "echo_generator", "director_remote"), |
| 283 | serialization_alias="echoGenerator", |
| 284 | ) |
| 285 | file_storage: FileStorageConfig = Field( |
| 286 | default_factory=FileStorageConfig, |
| 287 | validation_alias=AliasChoices("fileStorage", "file_storage"), |
| 288 | serialization_alias="fileStorage", |
| 289 | ) |
| 290 | memory_review: MemoryReviewConfig = Field( |
| 291 | default_factory=MemoryReviewConfig, |
| 292 | validation_alias=AliasChoices("memoryReview", "memory_review"), |
| 293 | serialization_alias="memoryReview", |
| 294 | ) |
| 295 | restrict_to_workspace: bool = False # restrict all tool access to workspace directory |
| 296 | mcp_servers: dict[str, MCPServerConfig] = Field(default_factory=dict) |
| 297 | ssrf_whitelist: list[str] = Field(default_factory=list) # CIDR ranges to exempt from SSRF blocking (e.g. ["100.64.0.0/10"] for Tailscale) |
| 298 | |
| 299 | |
| 300 | class PromptStackerConfig(Base): |
| 301 | """Prompt Stacker monitoring configuration.""" |
| 302 | |
| 303 | enabled: bool = False |
| 304 | max_traces: int = 100 |
| 305 | |
| 306 | |
| 307 | class PromptEngineeringConfig(Base): |
| 308 | """Runtime-switchable Prompt Engineering (PE) set configuration.""" |
| 309 | |
| 310 | enabled: bool = True |
| 311 | root: str | None = None # PE sets root dir; defaults to the repo-level ``pe/`` when unset. |
| 312 | active: str = "v7_cinematic_full" |
| 313 | |
| 314 | |
| 315 | class Config(BaseSettings): |
| 316 | """Root configuration for nanobot.""" |
| 317 | |
| 318 | agents: AgentsConfig = Field(default_factory=AgentsConfig) |
| 319 | channels: ChannelsConfig = Field(default_factory=ChannelsConfig) |
| 320 | providers: ProvidersConfig = Field(default_factory=ProvidersConfig) |
| 321 | api: ApiConfig = Field(default_factory=ApiConfig) |
| 322 | gateway: GatewayConfig = Field(default_factory=GatewayConfig) |
| 323 | tools: ToolsConfig = Field(default_factory=ToolsConfig) |
| 324 | prompt_stacker: PromptStackerConfig = Field( |
| 325 | default_factory=PromptStackerConfig, |
| 326 | validation_alias=AliasChoices("promptStacker", "prompt_stacker"), |
| 327 | ) |
| 328 | prompt_engineering: PromptEngineeringConfig = Field( |
| 329 | default_factory=PromptEngineeringConfig, |
| 330 | validation_alias=AliasChoices("promptEngineering", "prompt_engineering"), |
| 331 | ) |
| 332 | |
| 333 | @property |
| 334 | def workspace_path(self) -> Path: |
| 335 | """Get expanded workspace path.""" |
| 336 | return Path(self.agents.defaults.workspace).expanduser() |
| 337 | |
| 338 | def _match_provider( |
| 339 | self, |
| 340 | model: str | None = None, |
| 341 | *, |
| 342 | provider: str | None = None, |
| 343 | ) -> tuple["ProviderConfig | None", str | None]: |
| 344 | """Match provider config and its registry name. Returns (config, spec_name).""" |
| 345 | from nanobot.providers.registry import PROVIDERS, find_by_name |
| 346 | |
| 347 | forced = self.agents.defaults.provider if provider is None else provider |
| 348 | if forced != "auto": |
| 349 | spec = find_by_name(forced) |
| 350 | if spec: |
| 351 | p = getattr(self.providers, spec.name, None) |
| 352 | return (p, spec.name) if p else (None, None) |
| 353 | return None, None |
| 354 | |
| 355 | model_lower = (model or self.agents.defaults.model).lower() |
| 356 | model_normalized = model_lower.replace("-", "_") |
| 357 | model_prefix = model_lower.split("/", 1)[0] if "/" in model_lower else "" |
| 358 | normalized_prefix = model_prefix.replace("-", "_") |
| 359 | |
| 360 | def _kw_matches(kw: str) -> bool: |
| 361 | kw = kw.lower() |
| 362 | return kw in model_lower or kw.replace("-", "_") in model_normalized |
| 363 | |
| 364 | # Explicit provider prefix wins — prevents `github-copilot/...codex` matching openai_codex. |
| 365 | for spec in PROVIDERS: |
| 366 | p = getattr(self.providers, spec.name, None) |
| 367 | if p and model_prefix and normalized_prefix == spec.name: |
| 368 | if spec.is_oauth or spec.is_local or p.api_key: |
| 369 | return p, spec.name |
| 370 | |
| 371 | # Match by keyword (order follows PROVIDERS registry) |
| 372 | for spec in PROVIDERS: |
| 373 | p = getattr(self.providers, spec.name, None) |
| 374 | if p and any(_kw_matches(kw) for kw in spec.keywords): |
| 375 | if spec.is_oauth or spec.is_local or p.api_key: |
| 376 | return p, spec.name |
| 377 | |
| 378 | # Fallback: configured local providers can route models without |
| 379 | # provider-specific keywords (for example plain "llama3.2" on Ollama). |
| 380 | # Prefer providers whose detect_by_base_keyword matches the configured api_base |
| 381 | # (e.g. Ollama's "11434" in "http://localhost:11434") over plain registry order. |
| 382 | local_fallback: tuple[ProviderConfig, str] | None = None |
| 383 | for spec in PROVIDERS: |
| 384 | if not spec.is_local: |
| 385 | continue |
| 386 | p = getattr(self.providers, spec.name, None) |
| 387 | if not (p and p.api_base): |
| 388 | continue |
| 389 | if spec.detect_by_base_keyword and spec.detect_by_base_keyword in p.api_base: |
| 390 | return p, spec.name |
| 391 | if local_fallback is None: |
| 392 | local_fallback = (p, spec.name) |
| 393 | if local_fallback: |
| 394 | return local_fallback |
| 395 | |
| 396 | # Fallback: gateways first, then others (follows registry order) |
| 397 | # OAuth providers are NOT valid fallbacks — they require explicit model selection |
| 398 | for spec in PROVIDERS: |
| 399 | if spec.is_oauth: |
| 400 | continue |
| 401 | p = getattr(self.providers, spec.name, None) |
| 402 | if p and p.api_key: |
| 403 | return p, spec.name |
| 404 | return None, None |
| 405 | |
| 406 | def get_provider( |
| 407 | self, |
| 408 | model: str | None = None, |
| 409 | *, |
| 410 | provider: str | None = None, |
| 411 | ) -> ProviderConfig | None: |
| 412 | """Get matched provider config (api_key, api_base, extra_headers). Falls back to first available.""" |
| 413 | p, _ = self._match_provider(model, provider=provider) |
| 414 | return p |
| 415 | |
| 416 | def get_provider_name( |
| 417 | self, |
| 418 | model: str | None = None, |
| 419 | *, |
| 420 | provider: str | None = None, |
| 421 | ) -> str | None: |
| 422 | """Get the registry name of the matched provider (e.g. "deepseek", "openrouter").""" |
| 423 | _, name = self._match_provider(model, provider=provider) |
| 424 | return name |
| 425 | |
| 426 | def get_api_key( |
| 427 | self, |
| 428 | model: str | None = None, |
| 429 | *, |
| 430 | provider: str | None = None, |
| 431 | ) -> str | None: |
| 432 | """Get API key for the given model. Falls back to first available key.""" |
| 433 | p = self.get_provider(model, provider=provider) |
| 434 | return p.api_key if p else None |
| 435 | |
| 436 | def get_api_base( |
| 437 | self, |
| 438 | model: str | None = None, |
| 439 | *, |
| 440 | provider: str | None = None, |
| 441 | ) -> str | None: |
| 442 | """Get API base URL for the given model, falling back to the provider default when present.""" |
| 443 | from nanobot.providers.registry import find_by_name |
| 444 | |
| 445 | p, name = self._match_provider(model, provider=provider) |
| 446 | if p and p.api_base: |
| 447 | return p.api_base |
| 448 | if name: |
| 449 | spec = find_by_name(name) |
| 450 | if spec and spec.default_api_base: |
| 451 | return spec.default_api_base |
| 452 | return None |
| 453 | |
| 454 | model_config = ConfigDict(env_prefix="NANOBOT_", env_nested_delimiter="__") |
| 455 |