Module settings

Expand description

Settings domain types and validation.

This module contains the core settings types used across the application. These are pure domain types with no infrastructure dependencies.

Modules§

settings_loop_guard 🔒: What the proxy’s loop guard does when a replayed history trips it.
settings_remote 🔒: The remote tunnel’s settings (ADR 0012): the connect side’s stored pairing as one value, and the remote half of merging and validating.
settings_update 🔒: The two types that sit beside Settings: a partial update, and what validation refuses.
settings_validate 🔒: The inference-parameter validators, split from super.

Device: One device this machine has issued a key to.
RemotePairing: The machine gglib remote join paired with, and the key that machine issued — one record, because they are one fact.
RemoteServe: How this machine was told to put its proxy on the tunnel, kept so a restart arms it the same way.
Settings: Application settings structure.
SettingsUpdate: Partial settings update.

LoopGuardMode: What the loop guard does with a request whose replayed history trips it.
SettingsError: Settings validation error.

CONTEXT_SIZE_RANGE: The context sizes a person is allowed to configure.
DAEMON_PORT: Fixed loopback port for the gglib daemon’s management API.
DEFAULT_CONTEXT_SIZE: Default context size for models when not specified by the user.
DEFAULT_LLAMA_BASE_PORT: Default base port for llama-server instance allocation.
DEFAULT_PROXY_PORT: Default port for the OpenAI-compatible proxy server.
DEFAULT_REMOTE_PORT: The loopback port gglib remote join tries first for the paired machine.