Editing

Detailed explanation of the sections and functionalities in the editing page.

Instructions

This is a message sent to users which indicate what input is expected, and it will NOT be sent to the LLM.

Please ALSO include the variables mentioned here in the Prompt Body if need to be processed by LLM.

Edition Page

Prompt Body

You may noticed iSiri has already divide the body into different sections as you previous selection upon the Structure. You may also modify the sections and add your own section according to your creation.

The prompt body will be sent and processed by LLMs as system message when used by others.

Variables

You may add variables for your users to fill in / select, so that the prompt will be customized according to the users input.

If variables are added, they MUST show in the Instruction section for the user to fill in / select.

Use "/" to insert Variables into the Prompt body If the Prompt body does not contain variables, then users input will NOT be sent to the LLM, so it is highly recommended to include all the variables in the Prompt body.

Template

When you find useful prompt lines or text, it is highly recommended to store them in Template so that you can quote in other creation. The content from the Template will be sent to the LLM word by word.

For example, "You need to think step by step. Re-evaluate the answer before outputting" and "remember the user's characteristics, and important events and adjust according to user's requests" are some useful lines to store in the Template.

Click the pencil icon to open the text window for typing. You may create folders to store the lines. You have to create and select the the template before inserting it into the prompt.

When you wish to insert the template, click "+Add" and then select the record. You will see a new section in the prompt body with template inserted. You may adjust or insert the template else where using "/ "

Knowledge base

You can link the prompt with Knowledge base to store and manage external data, and enhance retrieval capabilities of created prompt/character for more informed and accurate responses (RAG, GraphRAG are currently supported). Check more details in Knowledge base.