@diverga/setup

Researcher onboarding and setup flows for Long Table.

Install

npm install @diverga/setup

CLI

Interactive:

diverga-setup init --flow interview

The interactive flow asks only for the global researcher profile and default challenge preferences. Project-specific goal and blocker intake should happen later in longtable start.

The interactive flow uses numbered choices by default. Fields that need custom text expose a None of the above option and then ask for free text.

Non-interactive:

diverga-setup init \
  --flow interview \
  --provider codex \
  --field education \
  --career-stage doctoral \
  --experience intermediate \
  --checkpoint balanced \
  --entry-mode explore \
  --panel-preference show_on_conflict \
  --write \
  --install \
  --json

Show a stored setup:

diverga-setup show --json

Install provider runtime config from stored setup:

diverga-setup install --json

By default this writes:

• setup output to ~/.diverga/setup.json
• Codex runtime config to ~/.diverga/runtime/codex/diverga.toml
• Claude runtime config to ~/.diverga/runtime/claude/diverga.json

The generated runtime config does not overwrite platform-native config files directly. It creates Long Table-managed runtime artifacts that can later be wired into provider-specific runtimes during migration.

Package Role

The setup contract combines technical setup with researcher-profile calibration rather than treating installation as a purely technical step.

Global setup should answer:

• who the researcher is
• how much challenge or slowdown they want
• what makes writing still feel like theirs
• how visible disagreement should be by default

Project and session intake belongs to longtable start, not diverga-setup init.

Included Outputs

• quick setup question flow
• provider selection resolution
• persisted setup output generator
• saved setup output helpers
• runtime config installer helpers
• numbered checkpoint helpers

See examples/ for sample Codex and Claude setup outputs.