Install¶

Scelo TUI installs from source. There is no release binary yet, and no build step either: bun link symlinks the live source, so edits take effect on the next run.

Prerequisites¶

Bun 1.3 or newer.
A terminal at least 140 columns wide. Below that it refuses to draw the three panes and says so, rather than rendering an unreadable mess. A portrait window can go narrower, since the panes stack: see The three panes.
The scelo repo checked out as a sibling. This is the part that catches people out, so it has its own section below.

The sibling checkout¶

@scelo/core is consumed over a file: path, ../scelo/packages/scelo-core, so bun install fails here without it. That is deliberate rather than an oversight: profiling, typing, coercion and filtering have exactly one definition, shared with the Scelo IDE, so the two projects cannot drift apart.

Clone them together:

git clone git@github.com:intelligentactuaries/scelo.git
git clone git@github.com:intelligentactuaries/scelo-tui.git
cd scelo-tui

If your scelo checkout lives somewhere else, point the file: path in package.json at it instead of moving the directory.

Install¶

bun install
bun link            # puts `scelo` on your PATH via ~/.bun/bin

bun link is what makes scelo runnable from any directory. Without it you can still run the app with bun run src/main.tsx, but relative paths will resolve against the repo rather than against wherever your data is.

Check it worked¶

scelo --help        # or just: scelo

You should land on the model picker. If instead you get

Terminal too small for the three panes.
100x30 — needs 140x24 side by side, or 56x32 stacked. Make it taller.

then the window is the problem, not the install. Widen it past 140 columns, or make it tall and narrow so the panes stack.

Optional, and worth having¶

	why
Ollama	the default model runs locally with no key and no cloud round-trip. `ollama pull qwen2.5:7b-instruct-q4_K_M` gets you the default.
R	so the exported `analysis.R` runs where you land. The generated script needs no packages, only base R.
Python with pandas	same, for `analysis.py`.
A Jupyter frontend	`/open notebook` looks for `jupyter-lab`, then `jupyter-notebook`, then bare `jupyter`.

None of these are required to start. Without a model the pipeline still ingests, profiles and cleans; the narrative and the bots go inert and say so. Without R, Python or Jupyter the export still writes every file, you just open them elsewhere.

Uninstall¶

bun unlink          # from inside the repo

Then delete the two checkouts. The only thing outside them is the config file at ~/.config/scelo-tui/config.json, which holds your model choice and any API keys. See Configuration.