Skip to content

Commands

Thirteen of them. None reach the model: they are actions rather than conversation, so they keep working when the model is down, slow or unconfigured.

Typing / opens a filtered menu of the lot. /help prints the same list in the pane.

command does
/files [folder] pick a data file to load, no dragging needed
/example [number|name] load a bundled sample dataset
/export [format…] write artifacts for every tool
/live [off] mirror the session into RStudio and Jupyter files as it runs
/open [format] open an exported artifact
/list the analyses that apply to this data
/charts [number] every plot this data makes, full screen
/run <analysis|number> switch the analysis
/show <column> one column's profile
/graph [on|off] the node and edge diagrams in TOOLS and HARD
/copy [table|reading|reply] put values on the clipboard, no dragging
/mouse [on|off] click-to-focus, at the cost of plain drag-select
/help everything you can type

Loading data

/files

/files [folder] opens a picker of the data files under the current directory, or under folder if given. Typing filters it; each entry shows its size and modification time.

data files under . — type to filter
▸ book.csv
  claims_2024.csv
  … 9 more
  11.5 KB · 15:20
  ↑↓ move · ⏎ pick · esc cancel

/example

/example opens a pick-list of the Scelo IDE's bundled samples. /example dirty loads one directly by name, and "load example data" as plain prose works too.

Reach for /example dirty to watch auto-clean do real work.


Choosing what runs

/list

The analyses that apply to this data, numbered. Opens as a modal sub-list: arrows move, Enter picks, typing filters, Esc backs out.

After it has printed, a bare number answers it.

/run

/run gini by name, /run 3 by the number /list printed. The HARD pane re-renders. The pipeline does not restart and the dataset is not re-read.

Needs an argument, so picking it from the / menu completes the line to /run and hands it back rather than submitting a usage error.

/show

/show premium prints that column's profile as a card: type, cardinality, missingness, range, distribution. The quickest check of the agent's reading against the actual data. Needs an argument.


Seeing it

/charts

/charts opens the full-screen gallery, /charts 3 opens it on plot 3. Arrows move, Q Esc or Enter returns to the panes.

/graph

/graph off turns off the node and edge diagrams in the TOOLS and HARD panes and gives those rows back to the tables and the chat. /graph on restores them. Bare /graph toggles.

/copy

Puts the values on the clipboard rather than a picture of them, with no box-drawing characters or column padding caught in the selection.

copies
/copy whatever the focused pane is showing
/copy table the result, tab separated, so it pastes into Excel as columns
/copy reading the agent's reading of the data
/copy reply the last thing the bot said

result and data are accepted as synonyms for table.

/mouse

/mouse on enables click-to-focus. /mouse off restores plain drag-select, which is the default. See Keys and mouse for the trade-off.


Getting it out

/export

/export writes everything; /export excel r writes just those formats. ctrl-e is the same as bare /export, as is clicking the header's ⇩ export tag. Where the files land depends on the host. Full detail in Export.

/live

/live arms the live mirror, rewriting R, notebook and CSV files at every milestone so RStudio or Jupyter can follow the session as it runs. /live off stops it, leaving the files in place.

/open

/open opens the export folder. /open [excel|python|notebook|r|sce|csv] opens one artifact, handing it to whatever should open that type rather than to the OS default in every case.

Works after /live with nothing exported yet, since the mirror's files stand in.


/help

Prints every command with its hint, plus the key bindings, built from the same registry the / menu reads so the two cannot drift apart.