n8n Nodes for AI Power Grid

Use AI Power Grid in n8n workflows for text, image, video, and audio generation. The node discovers currently advertised models instead of shipping a stale model catalog.

Credentials

Create a server-side API key in the Grid console. The key needs the account.read and inference.submit scopes carried by ordinary programmatic keys. account.read lets n8n validate the credential without spending credit; inference.submit authorizes generation. Add it to an AI Power Grid API credential in n8n. n8n stores the secret and adds it as a bearer token. The key is never a workflow field.

Do not put the key in browser code, workflow JSON, expressions, prompts, or screenshots. Give production workflows a separate, revocable key with only the scopes they need.

Operations

• Generate Text calls /v1/chat/completions and returns the completed OpenAI-style response. Standard n8n node executions do not relay SSE to the editor, so this operation deliberately uses stream: false.
• Generate Image calls /v1/images/generations for text-to-image. The first package does not claim img2img support because binary source handling needs a dedicated upload contract. Its model picker includes only workers advertising txt2img.
• Generate Video calls /v1/videos/generations for text-to-video. Its model picker excludes workers advertising only img2video.
• Generate Audio calls /v1/audio/generations for ACE-Step music, with optional lyrics and governed controls.

Image, video, and audio executions wait for the worker result and return the Grid response, including result URLs and provenance metadata. Their request timeouts follow Core's current modality deadlines. Use n8n queue mode and worker execution timeouts appropriate for long media jobs.

Example workflows

• Support draft: Manual Trigger -> AI Power Grid (Generate Text) -> Slack. Map the incoming support message into Prompt, then send choices[0].message.content to Slack for review.
• Campaign artwork: Schedule Trigger -> AI Power Grid (Generate Image) -> HTTP Request. Generate one image, then download the first returned result URL into n8n binary data for the next publishing step.
• Podcast bumper: Form Trigger -> AI Power Grid (Generate Audio) -> cloud storage. Map the submitted style notes into Music Description, use a short bounded duration, and store the returned audio URL.

Start with a Manual Trigger while building a workflow. Run one item, inspect the Grid response in the output panel, and map only the fields needed by the next node. Availability is dynamic, so select a currently advertised model instead of relying on an old saved model name.

Errors and credits

Grid HTTP errors are surfaced as n8n node errors. Common statuses are:

• 401: invalid or revoked API key
• 402: insufficient usable credits
• 404: model no longer available
• 422: request or model parameter is unsupported
• 429: rate or concurrency limit reached
• 503: no compatible worker is online

Fund the same Grid account at console.aipowergrid.io. Retry only operations your workflow can safely repeat; a media timeout does not prove that the remote job failed.

Privacy and positioning

Grid requests are routed to remote community-operated workers. Workers may be able to inspect plaintext prompts, source material, and outputs. Do not send secrets, personal or regulated data, or confidential source code unless a separately verified confidential-compute deployment satisfies your needs.

This package is maintained by AI Power Grid. It does not imply a partnership with or endorsement by n8n.

Development

Node.js 22 or newer is required.

npm ci
npm test
npm run lint
npm run build
npm pack --dry-run

A credentialed release check must exercise one bounded request in every operation before publication. Those checks spend account credit and are not part of the default test suite:

AIPG_LIVE_E2E=1 npm run test:e2e:live

Use a disposable account.read + inference.submit key and revoke it after the run. The live test does not print the key, prompts, generated media URLs, model output, or account balance.

Release

@aipowergrid/[email protected] is the current public package. Version 0.1.3 moves its npm provenance to this dedicated repository so n8n's Creator Portal can resolve credential source files at the repository root. Release it only through .github/workflows/publish.yml and the npm OIDC Trusted Publisher for this repository. The workflow rejects a tag that does not exactly match package.json, requires the release commit to be on main, and reruns the package checks before publishing. The one-time bootstrap token is revoked; do not restore a registry token.