Enterprise workflow integrations
Enterprise Intelligence provides a canonical versioned workflow API for definitions, approvals, runs, traces, human exceptions, delivery, and acceptance.
Build against the EI workflow APIAuraOne / Developers
Typed TypeScript and Python clients for evaluation workflows, signed webhooks, and REST endpoints for runs, evidence, and exports.
Explore our public TypeScript and Python SDKs, signed webhook specifications, and enterprise connector catalog below.
At a glance
Developer tools
Connect Enterprise Intelligence workflows, call the versioned evaluation API from TypeScript or Python, and verify signed webhook deliveries.
Enterprise Intelligence provides a canonical versioned workflow API for definitions, approvals, runs, traces, human exceptions, delivery, and acceptance.
Build against the EI workflow APICreate and track evaluation runs through versioned /api/v1 endpoints with typed request and response contracts.
Read the SDK referenceVerify the HMAC signature on each delivery before routing it into downstream systems. Event subscriptions are configurable per workspace.
Inspect event deliveryWork with reviewed clips. Each one keeps its session, its license scope, and its provenance as a single record.
Read the capture schemaTrace reviewed work from training through serving, rollback, and customer handoff.
Review infrastructureFirst commands
The quickstart switches language without changing the API contract or the evidence fields returned by the run.
npm install --save-exact @auraone/[email protected]import { AuraOneClient } from "@auraone/sdk";
// Platform-issued key: ak_<8 hex>_<64 hex>
const client = AuraOneClient.withApiKey(process.env.AURAONE_API_KEY!, {
// Optional: assert the organization — must match the key's org.
orgId: process.env.AURAONE_ORG_ID,
});
const run = await client.evaluations.create({
template_id: "YOUR_TEMPLATE_ID",
agent_bundle_url: "s3://YOUR_BUCKET/path/to/agent-bundle.zip",
config: { inputs: [{ prompt: "Hello", expected: "Hi" }] },
wait: false,
});
console.log(run.id, run.status);npm distribution. Current version: 1.0.0. Entry point: AuraOneClient, VERSION, DEFAULT_CONFIG.
Package documentationPyPI distribution. Current version: 1.0.0. Entry point: AuraOneClient, AsyncAuraOneClient.
Package documentationOperating trace
Follow each object from authentication through review, delivery, and downstream verification.
01
02
03
04
The coding sample
The sample ships with one script. It runs offline and needs no credentials. It tells you whether the files are the ones we shipped, and whether each task's recorded runs support its verdict. Node.js 20 or later.
node coding/run-all.mjsOptional: rebuild each task's starting code from its source. This needs git, and network access for tasks from public repositories. Nothing is installed and no test is run.
node coding/run-all.mjs --rebuild-start-trees --work /tmp/fs-replayOffline, from the files in the package.
The limits are in the package, and here.
Records you can read, and schemas that describe them.
There is no hosted grader, and no public access to hidden tests.
The same steps apply whether you start from an SDK, a REST endpoint, or a webhook.
01
Confirm the package, version, endpoint, event, and scope.
SDK manifest and API version
02
Use a source item that exercises the real criteria, review, and export path.
Request, response, and review result
03
Validate the signature, retries, idempotency, and archive destination.
Webhook and connector receipt
04
Record ownership, retention, failure handling, and the next release check.
Production ownership plan
Before launch
Confirm the interface, the authentication, and the recovery path before launch.