n8n-nodes-layahost

This is an n8n community node. It lets you use layahost in your n8n workflows.

layahost is a hosted typed-decision API: send a text and a question, and get back a structured answer (one option out of a list, yes/no with a probability, or a level on a scale) with a confidence, in a fraction of a second. You can also chain questions into decision flows in the layahost console and run them with one call.

n8n is a fair-code licensed workflow automation platform.

• Installation
• Credentials
• Operations
• Examples
• Using it as an AI agent tool
• Compatibility
• Resources

Installation

Follow the installation guide in the n8n community nodes documentation.

In short, on a self-hosted n8n: Settings → Community Nodes → Install, enter n8n-nodes-layahost, and confirm. Or from the command line in your n8n user folder:

cd ~/.n8n/nodes
npm install n8n-nodes-layahost
# restart n8n

Credentials

1. Sign up at layahost.com and create an API key on the API keys page. Keys start with lh_.
2. In n8n, create a Layahost API credential:
   • API Key: your lh_... key. It is sent as Authorization: Bearer <key>.
   • Base URL: leave https://layahost.com unless you run your own instance.
3. Click Test. n8n calls GET /v1/usage, which is free.

Operations

Decision → Decide

One text in, one decision out (POST /v1/decide).

• Text: the text to decide about, usually an expression such as {{ $json.body }}.
• Question Source:
  • Template: pick a ready-made question from the list (loaded from GET /v1/templates): moderation, spam, support intent, routing, sentiment, priority, language.
  • Custom Question: choose a Type and write the Question:
    • Choice: add 2 or more Answers, each with a name and an optional description.
    • Yes / No: no options needed.
    • Score: add 2 to 10 Levels, lowest first.
• Options (optional): model (laya-auto, laya-english, laya-multilingual), language hint, cache on/off, and a question override for templates.

Output: the API response, with decision at the top level:

{
  "decision": "billing",
  "confidence": 0.6506,
  "probabilities": { "billing": 0.9036, "technical": 0.0506, "sales": 0.0458 },
  "template": null,
  "model": "laya-english",
  "usage": { "decisions": 1, "input_tokens": 52, "cost_micros": 15, "cost_usd": 0.000015 },
  "latency_ms": 169.63,
  "request_id": "req_iazyijj7zo5nmmoxzhnivu9d"
}

• decision is the chosen option (choice), true/false (yes/no), or the level name (score).
• yes/no answers have probability (of yes) instead of probabilities; score answers also have score.

Decision → Decide Batch

The same decision for the text of every input item, in as few requests as possible (POST /v1/decide/batch, 64 texts per request). The question settings are read from the first item; the Text is read from each item. You get one output item per input item, in the same order, each with the same fields as Decide. A text that fails on its own (for example an empty one) fails only its item when Continue On Fail is on.

Batching saves round trips, not money: each text is billed as one decision.

Flow → Run

Runs a decision flow drawn in the layahost console (POST /v1/flows/{flow}/run).

• Flow: pick one from the list (loaded from GET /v1/flows). Each entry shows the flow's possible outcomes.
• Text: the text to run through the flow.
• Options → Language Hint (optional).

Output: the run, with outcome at the top level so a Switch node can branch on {{ $json.outcome }}:

{
  "flow": "support-inbox",
  "version": 1,
  "outcome": "billing-queue",
  "ended_at": "billing",
  "steps": [
    { "step": "spam", "label": "Spam?", "answer": "legit", "probability": 0.8049, "followed": "legit" },
    { "step": "intent", "label": "What do they want?", "answer": "refund", "probability": 0.9977, "followed": "refund" }
  ],
  "usage": { "decisions": 2, "cost_micros": 10, "cost_usd": 0.00001 },
  "request_id": "req_pebc6fszrqvqwhetxbxruttk"
}

Account → Get Usage

Balance, plan and usage totals of the account (GET /v1/usage), over the last Days (1 to 90). Free to call.

Errors

API errors are shown with layahost's own message, for example Invalid or revoked API key. [authentication_error] or options: A score needs 2 to 10 levels. [validation_error]. 402 means the balance does not cover the request.

With Settings → On Error → Continue (Continue On Fail), a failed item is passed on as { "error": "...", "statusCode": 402 } instead of stopping the workflow. For 429, 503 and 529, turn on Retry On Fail: those requests are not billed.

Examples

Route support emails with a flow

Gmail Trigger → Layahost (Flow → Run, flow: support-inbox, text: {{ $json.text }}) → Switch on {{ $json.outcome }}

Add one Switch rule per outcome (billing-queue, engineering-queue, page-on-call, discard, …). The flow list in the node shows every outcome the flow can return.

Moderate comments, with a human in the loop

Webhook → Layahost (Decision → Decide, template: Moderation, text: {{ $json.body.comment }}) → IF

• {{ $json.decision === 'toxic' && $json.confidence > 0.3 }} → hide
• otherwise → publish, or send low-confidence ones to a person.

Tag a spreadsheet of reviews in one go

Google Sheets (read rows) → Layahost (Decision → Decide Batch, custom Score question "How happy is the customer?", levels: angry, unhappy, neutral, happy, delighted, text: {{ $json.review }}) → Google Sheets (update rows with {{ $json.decision }})

Custom choice with descriptions

Custom Question, Type Choice, question Which team should answer this email?, options:

Using it as an AI agent tool

The node is marked usableAsTool, so you can attach it to an AI Agent node as a tool. Let the model fill the Text (and, for custom questions, the Question) with $fromAI(), and keep the template, flow and options fixed so the agent gets a fast, cheap and consistent classifier.

Compatibility

Built with @n8n/node-cli 0.49 against n8n-workflow 2.x. Tested with n8n 2.40. No runtime dependencies.

Resources

• n8n community nodes documentation
• layahost API docs: decide, templates, flows, batch, errors

License

MIT