Module stack

aprender

Module stack 

Source
Expand description

Sovereign AI Stack integration types (spec §9) Sovereign AI Stack Integration (spec §9)

Defines integration types for the broader Pragmatic AI Labs Sovereign AI Stack. The .apr format is standalone but interoperates with sibling tools:

  • alimentar (“to feed”): Data loading and transformation (.ald format)
  • pacha: Model registry with versioning and lineage
  • realizar (“to accomplish”): Pure Rust inference engine
  • presentar: WASM visualization and playgrounds
  • batuta: Orchestration and oracle mode

§Architecture

alimentar → aprender → pacha → realizar
    ↓           ↓          ↓         ↓
             presentar (WASM viz)
                   ↓
             batuta (orchestration)

§Design Principles

  • Pure Rust: Zero cloud dependencies
  • Format Independence: Each tool has its own binary format
  • Toyota Way: Jidoka, Muda elimination, Kaizen

Structs§

ComponentHealth
Individual component health
FormatCompatibility
Format compatibility information
InferenceConfig
Realizar inference endpoint configuration (spec §9.4.1)
ModelVersion
Pacha model version entry
StackHealth
Stack health status

Enums§

DerivationType
Model derivation types tracked by Pacha (spec §9.3.2)
HealthStatus
Health status levels
ModelStage
Pacha registry stage for model lifecycle (spec §9.3.1)
QuantizationType
Quantization types for model compression
StackComponent
Stack component identifiers