In LaminarFlow, Instructions have their own dedicated section, akin to Presets. Here, you can craft and fine-tune your Instructions, testing them thoroughly before integrating them into your Flows. This process is essential for gradually refining your concepts and ensuring they perform as expected.

What are Instructions?

An Instruction in LaminarFlow is structured with the following fields:

  • Name: A unique identifier you assign to each instruction.
  • Model: The specific AI model chosen for execution.
  • Query: The heart of the Instruction. This is where you pose your question to the AI.
  • Preset: Optionally, select one of your Presets to apply pre-configured settings, speeding up the setup process.
  • Additional Settings: These mirror the settings in Presets and can be customized on a per-Instruction basis.

Execution Modes

Instructions can operate in two distinct modes:

  • Queued Query: In this mode, your query is processed by the AI, and the response is provided upon completion.
  • Live Stream: Engage with the AI in real-time, watching your query being answered live.

A switch in the interface allows you to toggle between these two modes, providing flexibility in how you interact with the AI.

Example of an Instruction

Consider this example titled "Calculate the next day":


Calculate the next day


GPT-4 8K Latest


What is the day after today?    

In this instance, no predefined Preset was applied. As with Presets, you have the option to control AI behavior using five parameters: Frequency Penalty, Presence Penalty, Temperature, Max Tokens, and Top P.