n8n-nodes-typesafe-ai

September 17, 2026 · View on GitHub

An n8n community node for the TypeSafe AI System One API.

TypeSafe's System One models (Jev) answer small, typed questions about your data — yes/no probabilities, one-of-N choices, and scores against ordered levels — with calibrated probabilities your workflow can branch on directly. No prompt engineering, no free-text parsing.

Installation
Credentials
Operations
Output reference
Rate limits and retries
AI Agent tool
Resources
Author and support

Installation

The package is published on npm as n8n-nodes-typesafe-ai. Both methods below follow n8n's community nodes installation guide.

Install from the n8n GUI (recommended)

Only users with an Owner or Admin role can install community nodes on a self-hosted instance.

Go to Settings > Community Nodes.
Select Install.
Enter n8n-nodes-typesafe-ai in Enter npm package name. To pin a version, append it, for example n8n-nodes-typesafe-ai@0.1.0.
Agree to the risks of using community nodes: select I understand the risks of installing unverified code from a public source.
Select Install. n8n installs the node and returns to the Community Nodes list.

The TypeSafe AI node then appears in the node panel. Upgrades, downgrades and uninstalls are managed from the same Settings > Community Nodes page — see GUI installation.

Manual installation (self-hosted)

Use this if your instance runs in queue mode or you prefer installing from the shell.

Access your n8n shell (for Docker):

docker exec -it n8n sh

Create ~/.n8n/nodes if it doesn't already exist, navigate into it, and install the package:

mkdir -p ~/.n8n/nodes
cd ~/.n8n/nodes
npm i n8n-nodes-typesafe-ai

Then restart n8n. To upgrade later, run npm update n8n-nodes-typesafe-ai in the same directory and restart; to remove it, run npm uninstall n8n-nodes-typesafe-ai. Full details: Manual installation.

n8n Cloud users can install community nodes only after the package is verified by n8n; until then this package is available on self-hosted instances only.

Credentials

Create a TypeSafe AI API credential with your API key from the TypeSafe dashboard. The key is sent as an Authorization: Bearer header. Use the credential's Test button to confirm it works (it calls GET /v1/models).

Operations

All question operations share these fields:

Field	Description
State	The content to evaluate. Plain text, or a JSON object/array (a chat log, a record, your app's current state). Expressions such as `{{ $json }}` that resolve to an object are sent as JSON.
Instructions	The question the model should answer about the state.
Options → Model	`jev-latest` by default. Pin a versioned ID such as `jev-1.13.0` if you have tuned thresholds against it.
Options → State Format	`Auto` (default) sends JSON when the state parses as an object/array, otherwise text. `Text` and `JSON` force one behaviour.
Options → Simplify	On by default. Turn off to receive the raw API response body.

Question → Ask Yes/No

Returns the probability that the answer is yes (a noul).

Criteria (optional): True Means / False Means descriptions.

Example — State: Help! My payouts have been failing for 3 days. Instructions: Does this convey urgency?

{ "noul": 0.92, "model": "jev-1.13.0", "usage": { "input_tokens": 312, "output_tokens": 48 } }

Question → Ask Choice

Picks one option from a set you define and returns the full probability distribution plus a confidence value.

Choices: one row per choice (Value key + optional Description rubric).

Example — Instructions: Which team should handle this? Choices: billing, technical, sales

{
  "choice": "technical",
  "probabilities": { "billing": 0.08, "technical": 0.85, "sales": 0.07 },
  "confidence": 0.82,
  "model": "jev-1.13.0",
  "usage": { "input_tokens": 312, "output_tokens": 48 }
}

Question → Ask Score

Rates the state against ordered levels (lowest first) and returns a probability-weighted score that can land between levels.

Levels: at least two, each a short description, ordered from lowest to highest.

Example — Instructions: How frustrated is the customer? Levels: Calm, Frustrated, Very angry

{
  "score": 1.6,
  "legend": { "0": "Calm", "1": "Frustrated", "2": "Very angry" },
  "probabilities": { "0": 0.05, "1": 0.3, "2": 0.65 },
  "confidence": 0.78,
  "model": "jev-1.13.0",
  "usage": { "input_tokens": 312, "output_tokens": 48 }
}

Question → Evaluate Questions

Asks several questions about the same state in one request — the cheapest and fastest way to use TypeSafe. Define questions with the form (each with an ID, type, instructions and criteria) or paste the questions map as JSON exactly as described in the API reference. For yes/no questions the optional Criteria collection holds True Means / False Means, exactly as in Ask Yes/No.

{
  "answers": {
    "is_urgent": { "type": "noul", "noul": 0.92 },
    "department": { "type": "choice", "choice": "technical", "probabilities": { "...": 0 }, "confidence": 0.82 }
  },
  "model": "jev-1.13.0",
  "usage": { "input_tokens": 340, "output_tokens": 96 }
}

Model → Get Many

Lists the model names and aliases your account can use. One item per model: { name, description, release_date }.

Output reference

Operation	Simplified output
Ask Yes/No	`noul`, `model`, `usage`
Ask Choice	`choice`, `probabilities`, `confidence`, `model`, `usage`
Ask Score	`score`, `legend`, `probabilities`, `confidence`, `model`, `usage`
Evaluate Questions	`answers` (keyed by question ID, each with its `type`), `model`, `usage`
Model → Get Many	`name`, `description`, `release_date`

With Simplify off, every question operation returns the raw body: { "model", "answers": { "answer": { ... } }, "usage" }.

confidence (Choice and Score) summarises how concentrated the probability distribution is; see Confidence. A noul near 0.5 means yes and no are about equally likely.

Rate limits and retries

TypeSafe returns 429 Too Many Requests or 529 Overloaded when you exceed your limits or the service is busy. This node does not retry on its own. Enable Retry On Fail in the node's settings (with a wait of at least 1 second) so n8n backs off and retries.