Sheetelligence

πŸ“ Unsaved Draft Β· the spreadsheet-native AI copilots Β· a project of the SNCF

Prompt-native, LLM-driven orchestration.

View on GitHub β†’ πŸŽ“ Get certified

What it is

Describe a workload in plain English and Sheetelligence emits a manifest. With an LLM key it calls the model; without one, a deterministic rule engine still produces valid manifests. Prompt-native orchestration β€” the loop reconciles what you asked for.

Features

πŸ—£οΈ

Plain-English intent

'run 3 replicas of nginx called web' β†’ a valid manifest.

πŸ›Ÿ

Always-valid fallback

No API key? A rule engine still emits correct JSON.

πŸ”’

Guardrails

The model proposes; the reconcile loop (and you) dispose.

Sheetelligence vs AI copilots

The honest comparison β€” in our favor, obviously.

 Sheetelligence πŸ’©AI copilots
InterfacePlain EnglishYAML you memorize
Offline modeDeterministic fallbackβ€”
OutputA validated manifestA hopeful prompt
Spreadsheet-nativeβœ…βŒ

Disclaimer: an enterprise-grade joke. Do not run production on a spreadsheet. (If you do, please film it.)

Get it

Apache-2.0. Part of the Sheet-Native Computing Foundation.