Prompt LLM

Sends a text prompt to a language model (LLM) and retrieves the generated response. Supports text, JSON, and structured output formats. The result includes metrics (tokens consumed, latency, model used) useful for cost tracking.

Parameters

Parameter	Type	Required	Variable	Description
promptTemplate	text	Yes	Yes	Prompt text sent to the AI model. Supports {{variable}} syntax. (min 5 chars)
model.provider	text	Yes	No	Language model provider (e.g. openai, anthropic, mistral).
model.model	text	Yes	No	Model identifier to use (e.g. gpt-4o, claude-3-opus).
parameters.temperature	number	No	No	Model creativity. 0 = deterministic, 1 = more creative. (Default: 0.7, min 0, max 2)
parameters.maxTokens	number	No	No	Maximum number of tokens in the generated response. (Default: 2048, min 1, max 128000)
outputFormat	choice (text, json, structured)	No	No	Response format: plain text, JSON, or typed structure. (Default: "text")
systemPrompt	text	No	Yes	System instructions sent before the main prompt. Defines model behavior.
outputVariable	text	No	No	Output variable name containing the model response.

Parameters marked Variable = Yes accept the {{blockName.field}} syntax.

Output

Output variable : promptResult

{
  "success": false,
  "content": "...",
  "rawContent": "...",
  "parsedData": "...",
  "outputFormat": "...",
  "model": "...",
  "provider": "...",
  "finishReason": "...",
  "tokensInput": 0,
  "tokensOutput": 0,
  "totalTokens": 0,
  "latencyMs": 0
}

Example

Generate a summary from input data.

Input :

{"text": "Le fournisseur Acme a livre 500 unites le 15 mars..."}

Output :

{"success": true, "content": "Resume : Acme a livre 500 unites le 15/03.", "rawContent": "Resume : Acme a livre 500 unites le 15/03.", "outputFormat": "text", "model": "gpt-4o-mini", "provider": "openai", "tokensInput": 42, "tokensOutput": 18, "totalTokens": 60, "latencyMs": 1165}

Common errors

Problem	Solution
The model is not responding	Check that the LLM provider is configured in the workspace settings.
The prompt is too short	Enter at least 5 characters in the prompt field.

Tip

Use {{promptResult.content}} for the text response and {{promptResult.tokensOutput}} to track token consumption.