Parameters / Variables
Variables are terms extracted from and implied by the user's request. The variables you configure and add to nodes in your skill become the parameters that are used when executing the analysis Variables for the skill and overall assistant can be found in the upper right-hand panel when viewing a skill in skill studio. Select the icon with the sliders to open the variables panel. When setting up a variable there are several important fields to consider.
Name: This is the name you will see in the diagnostics for a skill or assistant. This is also the name that the LLM will see when assigning values. Try to make the name descriptive and clear - representative of the value you need to extract. Ex: If your metric variable is going to be a multi, go ahead and make it plural in the name (metrics) to help the LLM out.
Include as chat parameter: This toggle determines if the variable will be informed by the user's chat question. There is rarely a scenario where you would not want to set it as a chat parameter unless you want the variable to be ignored/excluded from the chat pipeline.
Limit acceptable user values with: This selection allows your to limit the types of values that should be inserted for this variable. If you have a particular data type in mind such as metric, it can vastly help in the parameter selection phase to have the options limited to "metrics"
LLM Description: This is the description that will be seen by the LLM when assigning values to parameters. Ensure this is either a very clear and accurate representation of the intended function for this variable OR leave it blank. If you leave it blank, the LLM will take its best pass at writing one for you.
Variable Type: This required setting determines whether the variable can be multiple distinct values (ei: Bud, Corona, Stella) or just a single value string (ei: "Bud" or "Customers over the age of 10")
Default: Here you can set the specific value(s) that you would like to use if the variable is not included in the user's query.
Here are some examples of variables...
- Metrics
- Dimensions
- Filters:
- Limit: (typically numbers from 1-X)
- Example Variable Description: this is a single numerical value representing the number of rows of data the user would like to analyze
- Sorting: Accepts dataset values, node values, or variables
- Example Variable Description: represents how the user would like the analysis to be sorted. use a format like {"col" : "column_name", "dir" : "asc or desc"}
Updated 22 days ago