Dailify
Log inJoin

AI

All AI templates share one node type on the canvas; the subtype picks the executor (generate text, summarise, classify, extract, chat, or transform). Each subtype has its own Execution fields and Output shape.

Overview

AI templates call a language model through the product's AI gateway. You choose a model, craft instructions (or subtype-specific bodies), and map structured outputs on the Output tab so downstream steps receive stable JSON keys.

Steps in this family

  • Generate text: open-ended generation from instructions plus optional tool-style behaviour configured in the sheet.
  • Summarise: condense supplied content; optional guidance steers format without overriding the core task.
  • Classify: choose exactly one label from a catalogue you define; confidence and reasoning fields map to exe outputs.
  • Extract: fill a typed field list from source content; enforced schema drives the model contract.
  • Chat: multi-turn style messaging with the model using thread context configured on the node.
  • Transform: rewrite or restructure prior content according to instructions and an optional dedicated content expression.

Configuration

Execution tab: model selector, instructions or optional guidance (depending on subtype), and any per-template bodies (for example content to summarise).

Output tab: declare outbound keys and map from {{exe.*}} fields after a test run or use import helpers where available.

Data flow

Instructions and content fields resolve expression variables from the inbound step, globals, constants, and time helpers. See Expression variables. Output mappings run after the model returns, in an extended context that includes exe.

Related