To do the same from a terminal, see Use of CLI Server.
Server
Use
Examples of what to ask your AI for what you want to do with published content, with the tools it uses and a few tips. The tools are what the AI uses; you do not call them yourself.
The AI calls search with scope: “all” and shows the matching data with its URL. To read one, it calls fetch with the id.
The AI calls fetch with the id it found through search and reads the body of the note or dashboard as Markdown. For a long note, it reads only the parts it needs with the section and query options of fetch. Along with the note name, tell it the topic you care about (for example “only the forecast section”) to read faster.
To do the same from a terminal, see Read notes.
Describe the data and your question. The AI checks the columns and aggregation rules before calculating. Review the calculation conditions (aggregation, filters) and the data date alongside the result.
To do the same from a terminal, see Query data.
The AI looks up the chart name in the dashboard and gets the numbers. For charts in notes, dashboards and slides, it can also filter, sort or aggregate them with chart_query. The numbers are the values as published.
To do the same from a terminal, see Get chart numbers.
The AI confirms the name of the analytics in the dashboard with fetch, then gets the result with analytics_get_result. Values such as coefficients and p-values come back as displayed. Tell it the name (title) of the analytics and the numbers you want (coefficients, p-values and so on).
To do the same from a terminal, see Get analytics results.
The AI filters or aggregates with the same conditions as query_dataset, saves the result as CSV or Parquet with dataframe_export, and returns a download link. Use it when the result is too large to get on screen. Tell it the format (CSV or Parquet) and the aggregation conditions. Data whose publisher does not allow downloading cannot be saved.
To do the same from a terminal, see Save results to a file.
