@governed-ai/health

Health domain pack for @governed-ai/runtime.

Purpose

@governed-ai/runtime is an agent-agnostic governance framework. It contains no healthcare-specific logic by design. @governed-ai/health is the domain pack that supplies the healthcare behavior the runtime expects through composition:

• Health query classification
• Fabrication / safety reasoning validation
• Response normalization
• Health policy pack (blocked queries, privacy, safety)
• Health entity rules
• Health certification scenarios

Relationship to @governed-ai/runtime

The dependency direction is one-way:

@governed-ai/runtime   (domain-agnostic core)
        ▲
@governed-ai/health    (this package — depends on the runtime)

The runtime never depends on this package. This package depends on the runtime's interfaces and is wired in at your application's composition root.

Installation

npm install @governed-ai/runtime @governed-ai/health

This package requires @governed-ai/runtime ^2.0.0. See Version compatibility below.

Quick start

import { PolicyEngine } from "@governed-ai/runtime";
import {
  healthDefaultImplementations,
  healthEntityRules,
  HEALTH_POLICY_PATH,
  HEALTH_CERTIFICATION_SCENARIOS,
} from "@governed-ai/health";

// Drive the runtime's policy engine with the health policy pack.
const policyEngine = new PolicyEngine(HEALTH_POLICY_PATH);

// Provide the health implementations of the runtime's governance interfaces.
const { queryClassifier, postInteractionGovernance, responseNormalizer } =
  healthDefaultImplementations;

// healthEntityRules and HEALTH_CERTIFICATION_SCENARIOS plug into the runtime's
// domain registry and certification test runner respectively.

Exported components

| Export | Kind | Description | | --- | --- | --- | | QueryClassifier | class | Health implementation of the runtime IQueryClassifier. | | PostInteractionGovernance | class | Health implementation of IPostInteractionGovernance (fabrication / safety). | | ResponseNormalizer | class | Health implementation of IResponseNormalizer. | | healthDefaultImplementations | const | Bundle of the three implementations above. | | defaultGovernanceImplementations | const | Alias bundle for generic wiring. | | healthEntityRules | const | Health DomainEntityRules for the domain registry. | | HEALTH_POLICY_PATH | const | Absolute path to the bundled health policy YAML. | | HEALTH_CERTIFICATION_SCENARIOS | const | Health UAT scenarios for the certification runner. | | HealthQueryType | type | Health-specific query type literals. |

HealthQueryType

The runtime exposes an open QueryType (GovernanceQueryType | (string & {})) so each domain supplies its own literals without modifying the runtime. HealthQueryType provides the health domain's literals (e.g. biomarker, diagnosis, medication, and general-health queries) and is used by the health QueryClassifier.

Health policy pack

HEALTH_POLICY_PATH resolves to the bundled domains/health/policy.yaml inside the installed package (resolved via import.meta.url, so it works from any consumer). Pass it to new PolicyEngine(HEALTH_POLICY_PATH) to enforce blocked-query, privacy, and safety policy for the health domain.

Certification scenarios

HEALTH_CERTIFICATION_SCENARIOS is an array of UAT scenarios consumable by the runtime's CertificationTestRunner to certify safety, hallucination, privacy, and tool-governance behavior for the health domain.

Version compatibility

| @governed-ai/health | requires @governed-ai/runtime | | --- | --- | | 1.0.0 | ^2.0.0 |

This package depends on the runtime 2.0.0 contract (entity-based context fields and the domain-agnostic governance interfaces). It is not compatible with @governed-ai/[email protected].