Pranor Learn — ML Inference Provider

Version: 2.0.0-dev
Module Path: github.com/vyuvaraj/pranor/learn/api
License: AGPL-3.0 (OSS) / EE

Overview

Pranor Learn acts as a pluggable ML inference provider for the Decision Engine, powering the Level 5 Learn veto level.

The module is divided into three sub-modules:

• learn/api: Shared interfaces and contracts.
• learn/wasm: Wazero runner for WASM inference.
• learn/sidecar: gRPC IPC sidecar for external models.

Zero-CGO Constraint

All Pranor Learn code enforces CGO_ENABLED=0. Complex ML inference is offloaded to the gRPC sidecar.

API Reference

Predictor Interface

type Predictor interface {
    Predict(ctx context.Context, in PredictInput) (PredictOutput, error)
    HealthCheck(ctx context.Context) error
}

Types

PredictInput Contains the features and context for inference, as well as BudgetMs for timeouts.

PredictOutput Contains the prediction result, confidence scores, and advisory actions.

Fault Contracts

• Returns ErrSidecarTimeout if the gRPC sidecar exceeds BudgetMs.
• Returns ErrModelBudgetExceeded for inference compute overruns.

Enterprise Edition