Packages

Agentic packages.

LuxoAI agents are built from small package contracts. Each package owns one part of the runtime: policy, chat, observability, feedback, or model operations.

Package map

Five packages share the runtime contract. Each owns one slice of the surface — policy, telemetry, chat, feedback, or model operations — and agents compose the pieces they need.

@luxoai-dev/agent-core

Agent Core

Defines agents, action registries, role policy, risk gates, and execution boundaries.

Open package

@luxoai-dev/observability

Observability

Service identity, health, readiness, metrics, and agent runtime signals.

Open package

@luxoai-dev/luxo-assistant

Luxo Assistant

Connects a registered agent to a bounded assistant surface inside the product.

Open package

@luxoai-dev/feedback-ui

Feedback UI

Collects structured 1-5 quality signals for review, ranking, and improvement loops.

Open package

luxo-ml

Luxo ML

Model contracts, registry, evaluation, and promotion rules for agent signals.

Open package

API Reference

Production-grade reference for every export of the four TypeScript packages. Function signatures, parameter tables, return types, and runnable examples — the same level a senior engineer or an LLM agent needs to write production code without reading source.

All packages

API Reference index

Browse the per-package references with module headers, signatures, types, and examples.

Open index

@luxoai-dev/agent-core

Agent Core API

Action registry, agent runtime, LLM gateway, judge, audit sinks, feedback validation, provider adapters.

Open reference