Variables

Variables are the mechanism by which data can be both updated by and revealed to the LLM.

• A variable must be defined before you can use it.

• To update the value of a variable, you expose the variable to the LLM for update at select points in your conversation process.

• To reveal the value of a variable to the LLM, embed the variable in the prompting. The inference engine will substitute the actual value of the variable when assembling the prompting when calling the LLM.

• Variables can be revealed to the LLM to ensure it understands the state of your Agent and its conversation.

• Variables can also be used to parameterise the conversation, allowing you to alter the behaviour of your agent according to information elicited from the user or provided by an external system when the interaction with the Agent is first initiated.