Transforms a_llm

LLM Call

Single-shot LLM completion. Pair with memory_key to persist the reply.

What does it do?

Runs a quick text transformation using AI, without trying to chat.

How does it work?

Receives text and precise instructions (e.g. "Translate this to English" or "Extract the shipping address from the following message"). Executes the task and returns the pure result.

Why use it?

To process text without an interactive chatbot. Use it to build automated flows (e.g. reading reviews and extracting sentiment) efficiently.

Inputs

name	type
`text`	string

Outputs

name	type
`content`	string
`model`	string
`usage`	object
`finish_reason`	string

Branches

Fixed branches

out — Output

Configuration

model string optional

messages array<object> required default: []

input_field string required default: "text"

system_prompt string optional

temperature number required default: 0.7

max_tokens integer required default: 1024

memory_key string optional

memory_mode "replace" | "append" required default: "replace"

parse_json boolean required default: false

response_schema object optional

Default configuration

{
  "model": "deepseek/deepseek-v4-flash",
  "messages": [],
  "input_field": "text",
  "system_prompt": "",
  "temperature": 0.7,
  "max_tokens": 1024,
  "parse_json": false,
  "memory_mode": "replace"
}

Emitted events

llm.reply
LLM response with model, token usage and finish reason.
llm.validation_failed
LLM response did not match the configured response schema.