Polyglot Workflows

Polyglot Workflows

Block is most useful when each stage uses the language that already fits the task.

<py>
rows = [10, 20, 30]
result = {"count": len(rows), "total": sum(rows)}
</py>

<js>
console.log(`count=${result.count}, total=${result.total}`)
</js>

<json>
{
  "result": {{result}}
}
</json>

Design rules

1. Keep each runtime block focused on one responsibility.
2. Pass plain data at the boundary.
3. Name the values that downstream stages need.
4. Run block check before executing a workflow.
5. Keep secrets and live connections inside the stage that owns them.

When Block is valuable

Block is a good fit for local ETL, report generation, automation, teaching, and AI-assisted workflows where the complete sequence should remain visible in one file.

It is not intended to hide arbitrary shell commands or replace a mature runtime's package ecosystem.