Templates and lessons

Studio ships with Jobs that actually run. Opening one and reading its JSON is usually faster than reading a reference page.

Templates

Templates in the sidebar. Each card carries what the template demonstrates; Preview shows the JSON before you commit to anything, Use template copies it into a Workflow as a new Job.

Template	Level	What it teaches
CSV to Parquet	starter	The four transformations you use daily, and why order matters
Ingestion with data quality	starter	Validations, on_failure: warn, and a persisted report
Join and runtime pushdown	intermediate	checkpoint → collect → {{variable}} inside with_transformations
Nested payload and multi-output	intermediate	struct dot-paths, two destinations from one DataFrame
Delta merge (upsert)	intermediate	mode: merge, merge_keys, and the uniqueness rule that guards it
Aggregation with group_by and pivot	intermediate	Complete SQL aggregate expressions, listed pivot values
Union and dedupe	intermediate	Positional vs by-name union, deduplication
Staging view handoff	intermediate	Several Jobs composing through a temp view — the shape a Pipeline automates
Kafka publication	advanced	struct → to_json → topic, as a per-destination transformation
Parameterized pipeline	advanced	{param} and skip_if_false branching

Every template is real Sparquet JSON, imported through the same compiler as any other file — and covered by a test that lints all of them without a single error.

Lessons

Learn in the sidebar holds six lessons that build on each other. Each has sections, runnable snippets and a checklist that remembers what you ticked.

1. Your first job in five minutes — the shape of the document and the run
2. Thinking in nodes — how the graph maps onto the file, including forks and joins
3. Transformations that matter — select vs with_column vs struct, and where each one wins
4. Joining and enriching — join semantics, and pushdown with collect + {{variable}}
5. Data quality — validators, on_failure policies, and reading the report
6. Going to production — parameters, includes, checkpoints, merge writes, and running on a cluster

Lessons link to the template that demonstrates them, so reading and doing stay one click apart.

Starting from a template in practice

The fastest path to a new Job is rarely a blank canvas:

1. Find the template closest to your case.
2. Use template into the right Workflow.
3. Change the source and destination paths.
4. Read the Issues panel — it will point at anything the swap invalidated.
5. Run it with the local runner against a small slice of data.

For anything genuinely new, describe it to the AI assistant and edit what it drafts.