System One

Ask typed questions (choice, score, yes/no) about a text or data, or route it by choice, with calibrated confidence

Actions

Steps this node can perform. Add one from Add a step, then pick the action.

Ask Questions

Ask typed questions (choice, score, yes/no) about a text or data

Settings

Field	Type	Required	Description
Model	Choice	Yes	System One model. jev-latest moves with new releases; pin a version to keep answers stable Options: TypeSafe Jev (latest), TypeSafe Jev (preview), TypeSafe Jev 1.13.0 (pinned) Default: typesafe/jev-latest
State Source	Choice	Yes	Where the text or data the questions are about comes from Options: Text (expression), JSON Default: text
State (Text)	Text		The text to read, e.g. the message or document. Use an expression to pull it from an earlier node
State (JSON)	Text		A JSON object or array to read
Define Questions	Choice	Yes	Build the questions with the editor, or paste TypeSafe's questions map Options: Using Fields Below, Using JSON Default: fields
Questions	List		Questions to answer. Each ID becomes a field in the result. Switch to Expression to map questions from an earlier node
Questions (JSON)	Text		TypeSafe questions: a map {"is_relevant": {"type": "noul", "instructions": "..."}} or an array with a key on each item. type is noul, choice or score; choice criteria map up to 255 options to a description or null; score criteria list 2 to 10 levels, lowest first; noul criteria {"true": ..., "false": ...} are optional and need both keys
Simplify Output	Yes / No		One flat value per question (plus _confidence and _level fields) instead of the full reply with probabilities and usage Default: true
Max Retries	Number		Retries on rate limiting (429) or overload (529) and other transient errors, with backoff that honors Retry-After Default: 3
Timeout (Ms)	Number		How long to wait for one response before failing the attempt Default: 60000
Custom Tags	List		Optional custom tags for AI Gateway cost tracking

Output

Later steps use this data by picking this step in the $ picker. For example, jev._model is $.<step ID>.output.jev._model.

{
  "jev": {
    "_model": "string"
  }
}

Example values show the shape of the data, not real values. See Using data from earlier steps.

Field descriptions

Field	Type	Description
jev	Object	The result under the output field (jev by default). With Simplify Output on: <id> (choice: option, score: weighted score, noul: probability of yes), <id>_confidence (choice, score), <id>_level (score) and _model. Off: answers, model_used, usage
jev._model	Text	Versioned model that answered

Route by Choice

Pick one route for a text or data. Branch on the result with a Path node: route names the choice, lowConfidence flags an unsure one

Settings

Field	Type	Required	Description
Model	Choice	Yes	System One model. jev-latest moves with new releases; pin a version to keep answers stable Options: TypeSafe Jev (latest), TypeSafe Jev (preview), TypeSafe Jev 1.13.0 (pinned) Default: typesafe/jev-latest
State Source	Choice	Yes	Where the text or data the questions are about comes from Options: Text (expression), JSON Default: text
State (Text)	Text		The text to read, e.g. the message or document. Use an expression to pull it from an earlier node
State (JSON)	Text		A JSON object or array to read
Instructions	Text	Yes	What to decide, e.g. What should be decided about this text? State boundary cases in the route descriptions; Jev reads literally
Routes	List	Yes	The choices. Add at least two. Route the result with a Path node on route
Low Confidence Handling	Choice	Yes	What to do when Jev is unsure Options: Flag as Low Confidence, Send to Best Route Anyway Default: low_confidence_output
Confidence Threshold	Number		Results with confidence below this set lowConfidence to true. Use a higher value when a wrong route is costly Default: 0.5
Max Retries	Number		Retries on rate limiting (429) or overload (529) and other transient errors, with backoff that honors Retry-After Default: 3
Timeout (Ms)	Number		How long to wait for one response before failing the attempt Default: 60000
Custom Tags	List		Optional custom tags for AI Gateway cost tracking

Output

Later steps use this data by picking this step in the $ picker. For example, jev._model is $.<step ID>.output.jev._model.

{
  "jev": {
    "_model": "string",
    "confidence": 0,
    "lowConfidence": false,
    "probabilities": {},
    "route": "string"
  }
}

Example values show the shape of the data, not real values. See Using data from earlier steps.

Field descriptions

Field	Type	Description
jev	Object	The route result under the output field (jev by default)
jev._model	Text	Versioned model that answered
jev.confidence	Number	Confidence in the chosen route, 0 to 1
jev.lowConfidence	Yes / No	True when confidence is below the threshold and low confidence handling is on
jev.probabilities	Object	Probability of each route
jev.route	Text	Name of the chosen route