Skip to main content

Module config

Module config 

Source
Expand description

Unified configuration system for Hyprstream

This module provides a layered configuration architecture:

  • HyprConfig: Root configuration combining all subsystems
  • ServerConfig: HTTP server configuration (network, CORS, TLS)
  • Model and runtime configs for ML inference

Re-exports§

pub use server::CorsConfig;
pub use server::NinepCanonicalReadConfig;
pub use server::SamplingParamDefaults;
pub use server::ServerConfig;
pub use server::ServerConfigBuilder;

Modules§

server
Server configuration with builder pattern

Structs§

At9pRenewalConfig
In-daemon at9p service-identity renewal ([registry.at9p_renewal]).
ClaimMapping
Claim field name overrides for mapping a userinfo JSON response to standard claims.
CredentialsConfig
Credentials storage configuration.
DiscoveryServiceConfig
Discovery service configuration.
EventServiceConfig
Event proxy service configuration.
FederateAtprotoOwnerEnrollment
Trusted deployment mapping from one Federate ATProto owner identity to an already-enrolled local hosted account. Both values come from launcher or operator configuration; neither is accepted from callback parameters.
FederateClientClassConfig
A second accepted Federate client class (W1: the client-side TUI harness).
FlightConfig
/// Arrow Flight SQL server configuration
GenerationConfig
Text generation parameters
GenerationResult
Generation result with metadata
HyprConfig
Unified configuration for the Hyprstream system
HyprConfigBuilder
Builder for Hyprstream configuration
HyprstreamTrainingConfig
Model-level training mode configuration (embedded in config.json under “hyprstream_training”)
InferenceServerConfig
Configuration for one deployable, CPU-only inference process.
LoraAppConfig
Application-level LoRA settings (TOML config: enabled, max_adapters, etc.)
MCPConfig
MCP service configuration (Model Context Protocol)
MacBootPolicyConfig
Both independently provisioned artifacts and their expected policy pin are mandatory when configured. Keep this generation/hash pin in the OS-owned deployment configuration, outside the replaceable artifact directory. Deploy a new pin and matching artifacts as one release; a mismatch fails startup closed. Recovery restores the matching backed-up config and artifacts. Lowering the pin is an explicit rollback decision. Trust keys are never supplied by these files or this configuration.
MeshPeerConfig
Configuration for a trusted mesh peer’s post-quantum signing identity (#157).
MetricsConfig
Metrics service configuration (DuckDB-backed time-series ingest + DataFusion query).
ModelConfig
Model loading and identification
OAIConfig
OpenAI-compatible HTTP API configuration
OAuthConfig
OAuth 2.1 authorization server configuration
OidcProviderConfig
Configuration for an external OIDC provider (login delegation).
PolicyServiceConfig
Policy service configuration.
QuicConfig
RdsConfig
Records/repository RDS (Multi-AZ Postgres) configuration.
RegistryServiceConfig
Registry service configuration.
ResolvedSamplingParams
Resolved sampling parameters with concrete values (no Options for required fields)
RpcServerConfig
StreamService configuration
RuntimeConfig
Runtime execution configuration
SamplingParams
Unified sampling parameters with Option fields for clean precedence merging.
SecretsConfig
Persistent secrets storage configuration.
ServicesConfig
Service management configuration
StorageConfig
Storage paths and directories configuration
StreamingConfig
moq streaming plane configuration.
TTTMetrics
TTT adaptation metrics (mirrors training::ttt::TTTResult)
TTTTrainingConfig
TTT-specific configuration
TemplatedPrompt
A prompt string that has been processed through the chat template engine. This newtype prevents accidentally passing untemplated strings to generation.
TlsConfig
TokenConfig
JWT token issuance configuration
TrustedIssuerConfig
Configuration for a trusted external OIDC issuer.
TuiServiceConfig
TUI display server configuration.
ValkeyCredentialsConfig
Valkey connection settings (used when backend = "valkey").
XetConfig
XetService configuration — HuggingFace-XET CAS HTTP face.

Enums§

CredentialsBackend
Which backend stores account credentials.
FinishReason
Why generation stopped
NativeNetworkProfile
QUIC/WebTransport transport configuration.
ProviderKind
Protocol kind for an external OAuth/OIDC provider.
ProvisioningMode
Whether to auto-create users on first external login.
TlsMode
TLS configuration for HTTP services (OAI, OAuth, MCP).
TrainingMode
Training mode configuration
UserMappingStrategy
How to map an external OIDC identity to a local hyprstream subject.

Constants§

MAX_SERVICE_IDENTITY_TTL_SECS
Maximum accepted service-identity lifetime: 90 days. The at9p chain layer has no maximum TTL of its own — this ceiling is provisioner-side policy. 90 days satisfies the anchor-TTL ≥ 3× worst-case-deploy-gap policy for the 30-day delegated-signer ceremony cadence, and the in-daemon renewal loop (services::at9p_renewal) renews at the half-TTL point so a live daemon never approaches it.
MIN_SERVICE_IDENTITY_TTL_SECS
Minimum accepted service-identity lifetime: shared by the offline provisioner’s --valid-for-seconds clamp and the in-daemon renewal TTL.

Functions§

default_lora_rank
default_target_modules
explicit_config_path
The canonical explicit selector, if the operator supplied one.
install_explicit_config_path
Install the canonical explicit config selector provenance (write-once).
install_pinned_config
Install the process-wide validated configuration snapshot (write-once).
pinned_config
The installed validated configuration snapshot, if startup pinned one.