This document explains how to use the Context feature. Jitera provides the Context function as a way to consolidate tacit knowledge that often appears in business operations.
Context is a feature that allows agents to memorize tacit knowledge. Tacit knowledge includes skills that are unique to individuals, meeting minutes that accumulate daily, and various other types of information. By memorizing this information, you can create agents specialized for your business operations.
By teaching agents the perspectives needed for hearings, you can create agents that check for missing points during interviews. By memorizing perspectives based on past hearing content, you can improve the thoroughness of your interviews.
By memorizing meeting minutes, you can create agents that remember the content of past meetings. Since it is difficult for humans to remember everything, teaching agents scattered information enables centralized management.
For example, you can instruct agents to always refer to @Coding Guidelines when generating code. You can also define agent behavior by referencing documents within context instructions. Other use cases include instructing agents to only output information that can be clearly read from code when generating design documents from code, thereby suppressing hallucinations.
Ask the agent about knowledge not held by generative AI and check the response when there is no information in the context.
Select an agent and enter the following prompt in the chat box:
"Jitera does not have information about itself, so a web search will be executed. From this state, you will teach Jitera’s information to the context."
After selecting an agent from the left menu, click "View Context" at the top of the screen.
Enter the following prompt in the message box at the bottom left of the context screen and click the execute button. (Before clicking, make sure "Learning Mode" is displayed in the chat box.)