Skip to main content

LLM node

This document describes how to create and use a Large Language Model (LLM) node .

The flow orchestration feature has been upgraded to workflow applications and agent orchestration applications to support a wider range of orchestration needs. The flow orchestration feature will be gradually phased out. This change does not affect current users. You can continue to use the flow orchestration feature or switch to workflow applications or agent orchestration applications for an improved experience. For more information, see workflow applications and agent orchestration applications.

Node interface

Go to Application Center > Flow Orchestration > New Flow, and then click Large Model Node.
image

Node description

The LLM node is a large model node. You can use this node to associate with models, such as official models and Supervised Fine-Tuning (SFT) models, to leverage their capabilities.

Configuration

In the large model node, you can select an official, third-party, or your own SFT model. In the text box below, enter a prompt. Prompts support dynamic variable replacement with variables from any part of the flow. For example, a variable such as ${xxx.yyy} is replaced when the flow runs. This provides the large model with a complete prompt.

Methods for entering variables

(1) Manual input

You can obtain the variable from the corresponding result based on its type. For example, ${bizVars.abc}. For more information, see Flow variable definitions and value retrieval.

(2) Suggestions

If the LLM node is connected to a parent node, enter / in the prompt text box. A list of variables appears. You can select a variable from the list. This has the same effect as entering the variable name manually.
image

(3) Loop configuration

The large model node supports loop configurations. For more information about the configuration, see Use loop properties.
Token Plan
Model Playground
Fine-tuning
Model Compression
Statistics and Monitoring
Support