Stats
Actions
Tags
From arelis-sdk
Generates and reviews code using the Arelis AI Governance SDK for TypeScript (@arelis-ai/ai-governance-sdk) and Python (ai-governance-sdk). Covers createArelis / create_arelis orchestration, governedInvoke / governed_invoke, agents.run, governance gates, managed PII config, platform events, causal graphs, policy engines, audit sinks, MCP, RAG, memory, quotas, and compliance. Triggers on: imports from @arelis-ai packages, from arelis import, GovernanceContext, createArelis, create_arelis, createArelisClient, create_arelis_platform, withGovernanceGate, with_governance_gate, governedInvoke, governed_invoke, or AI Governance SDK questions.
How this skill is triggered — by the user, by Claude, or both
Slash command
/arelis-sdk:ai-governance-sdkThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Governed AI orchestration framework supporting **TypeScript** and **Python**. Every operation needs a **GovernanceContext**, emits **audit events**, and integrates with the **Arelis Platform** for risk, compliance, and causal lineage.
references/python/agent-tool-patterns.mdreferences/python/api-reference.mdreferences/python/examples/agents-run.mdreferences/python/examples/governance-gate.mdreferences/python/examples/governed-invoke.mdreferences/python/examples/graphs-proofs-pipeline.mdreferences/python/examples/mcp-quotas-errors.mdreferences/python/examples/pii-and-policy.mdreferences/python/examples/platform-events-risk.mdreferences/python/examples/policy-crud.mdreferences/python/examples/setup-and-registration.mdreferences/python/governance-patterns.mdreferences/python/model-patterns.mdreferences/python/platform-pipeline.mdreferences/python/setup-patterns.mdreferences/python/testing.mdreferences/shared/concepts.mdreferences/shared/platform-api.mdreferences/shared/policies.mdreferences/typescript/agent-tool-patterns.mdGoverned AI orchestration framework supporting TypeScript and Python. Every operation needs a GovernanceContext, emits audit events, and integrates with the Arelis Platform for risk, compliance, and causal lineage.
@arelis-ai/ai-governance-sdk, uses createArelis, governedInvoke, withGovernanceGate, ArelisPlatform, .ts/.tsx filespip install ai-governance-sdk, imports from arelis (e.g. from arelis import create_arelis, GovernedInvokeInput), .py files, FastAPI/Django/Flask| Aspect | TypeScript | Python |
|---|---|---|
| Package | @arelis-ai/ai-governance-sdk | ai-governance-sdk (PyPI), import as from arelis import ... |
| Recommended entrypoint | createArelis() | create_arelis() |
| High-level model calls | arelis.governedInvoke() | arelis.governed_invoke() |
| High-level agent loop | arelis.agents.run() | arelis.agents.run() |
| Managed PII config | arelis.governance.getPiiConfig() | arelis.governance.get_pii_config() |
| Governance gate | withGovernanceGate() | with_governance_gate() |
| PII scanning | scanPromptForPii() | scan_prompt_for_pii() |
| Platform client | ArelisPlatform / arelis.platform | create_arelis_platform() / arelis.platform |
| Local governance client | createArelisClient() | Not available |
| Policy engine | Local PolicyEngine with checkpoints | Platform-side evaluate_policy() + managed PII |
| Audit sink | Local sink + platform events | Platform events only |
import { createArelis, type GovernedAgentTool } from '@arelis-ai/ai-governance-sdk';
const arelis = createArelis({
platform: {
apiKey: process.env.ARELIS_API_KEY!,
...(process.env.ARELIS_API_URL ? { baseUrl: process.env.ARELIS_API_URL } : {}),
},
aiSystemId, // optional: auto-propagated to all SDK surfaces
});
const result = await arelis.governedInvoke({
model: 'gemini-2.5-flash',
prompt: 'Summarize AI governance controls.',
denyMode: 'return',
invoke: async (sanitizedPrompt) => callModel(sanitizedPrompt),
});
from arelis import create_arelis, GovernedInvokeInput
arelis = create_arelis({
"platform": {
"apiKey": os.environ["ARELIS_API_KEY"],
**({"baseUrl": os.environ["ARELIS_API_URL"]} if os.environ.get("ARELIS_API_URL") else {}),
},
"aiSystemId": ai_system_id, # optional: auto-propagated
})
result = await arelis.governed_invoke(GovernedInvokeInput(
model="gemini-2.5-flash",
prompt="Summarize AI governance controls.",
invoke=lambda sanitized: call_model(sanitized),
deny_mode="return",
))
| Field | Type | Required | Description |
|---|---|---|---|
org | { id, name } | yes | Organization |
actor | { type, id, email?, roles? } | yes | Who — human | service | agent |
purpose | string | yes | Why — e.g. customer-support, chat |
environment | string | yes | Where — dev | staging | prod |
session_id | string | no | Session grouping |
tags | dict/object | no | Arbitrary key-value tags |
Optional aiSystemId set at config level is auto-forwarded through all platform-managed surfaces: governedInvoke, agents.run, governance gate telemetry, events.create, evaluatePolicy, risk.evaluate, proofs.create, and MCP evaluations.
Precedence: per-call > createArelis({ aiSystemId }) > platform: { aiSystemId } > omitted.
Compatibility: evaluatePolicy retries once without aiSystemId on HTTP 400 for backward compat.
@arelis-ai/ai-governance-sdk; named exports only, type imports for typescreateArelis({ platform, aiSystemId }) — SDK auto-forwards to all surfacesArelisPlatform base URL defaults to https://api.arelis.digitalwithGovernanceGate accepts ArelisPlatform directly; gate decisions include timingsresult.warnings; always await platform callscreateCompositeSink takes an array — NOT spread argsArelisClient has NO client.policy — export PolicyEngine directly for custom checkpointspip install ai-governance-sdk (NOT arelis); import from areliscreate_arelis + governed_invoke — handles PII, policy, events, risk automaticallygoverned_invoke accepts sync or async invoke; deny_mode="return" (default) or "throw"await; use try/except to swallow errorssource=arelis.platform (not arelis) for gate functions"allow", "deny", "warn", "escalate"startCausalGraph() BEFORE graphs.commit()| Task | Reference | Examples |
|---|---|---|
| Installation, create_arelis, web frameworks | setup-patterns.md | examples/setup-and-registration.md |
| governed_invoke, Gemini/Claude/OpenAI | model-patterns.md | examples/governed-invoke.md |
| PII scanning, BeforeToolCall/AfterToolResult | governance-patterns.md | examples/pii-and-policy.md |
| Governance gates (standalone + manual) | governance-patterns.md | examples/governance-gate.md |
| agents.run, governed agent loop | agent-tool-patterns.md | examples/agents-run.md |
| Platform policy CRUD | governance-patterns.md | examples/policy-crud.md |
| Events, evaluatePolicy, risk | platform-pipeline.md | examples/platform-events-risk.md |
| Causal graphs, proofs, post-stream pipeline | platform-pipeline.md | examples/graphs-proofs-pipeline.md |
| MCP, quotas, error handling | api-reference.md | examples/mcp-quotas-errors.md |
| Testing with mocks | testing.md | — |
npx claudepluginhub arelis-ai/ai-governance-plugin --plugin arelis-sdkAudits AI agent and LLM codebases for regulatory compliance (EU AI Act, GDPR, NIST AI RMF, HIPAA, ISO 42001). Scans for gaps, cross-references Arize instrumentation, and produces a remediation checklist.
Guides AI governance and compliance including EU AI Act risk classification, NIST AI RMF assessments, responsible AI principles, ethics reviews, and regulatory requirements for AI systems.
Conducts AI governance and responsible AI assessments using EU AI Act and NIST AI RMF, with risk classification, compliance evaluation, ethical reviews, and remediation roadmaps.