Penguin classifier

This example teaches the core-node path with the Palmer penguins data. It is small enough to read in one sitting and still shows the essential DAG pattern: source data, narrow the cohort, summarize the result, then build a report.

What it Teaches

• source loads a CSV with inferred column types.
• filter keeps one species.
• aggregate groups the filtered rows by island.
• rime build renders the node outputs as an HTML report.

Project Layout

Use the checked-in single-file example:

examples/single-file/
├── data/
│   └── penguins.csv
└── pipeline.dag.yaml

The fixture-level penguin parity test under packages/core/test/fixtures/experiments/penguin/ uses the same idea in its smallest form: penguins_source -> adelie_only.

DAG

specification_version: "2.1"

nodes:
  - id: penguins
    kind: source
    path: data/penguins.csv

  - id: adelie_only
    kind: filter
    inputs: [penguins]
    expr: '[species] == "Adelie"'

  - id: by_island
    kind: aggregate
    inputs: [adelie_only]
    groupBy: ["[island]"]
    metrics:
      - "[mean_bill_length] = [bill_length_mm].mean()"
      - "[mean_flipper_length] = [flipper_length_mm].mean()"
      - "[n] = [bill_length_mm].count()"

Run It

git clone https://github.com/danielsjoo/rime
cd rime
rime validate examples/single-file/pipeline.dag.yaml
rime run examples/single-file/pipeline.dag.yaml
rime build examples/single-file/pipeline.dag.yaml

Expected Output

by_island/default.parquet contains:

island	mean_bill_length	mean_flipper_length	n
Biscoe	40.30	195.00	1
Dream	39.15	180.00	2
Torgersen	39.30	183.50	2

Extend It

After the first run, try these small edits:

• Add metadata.report: false to the raw penguins source so only the filtered and aggregate nodes appear in the report.
• Add a derive node before by_island to compute a simple ratio:

- id: bill_to_flipper
  kind: derive
  inputs: [adelie_only]
  as: bill_to_flipper
  expr: "[bill_length_mm] / [flipper_length_mm]"

• Change the aggregate metrics to summarize the new column:

metrics:
  - "[mean_ratio] = [bill_to_flipper].mean()"
  - "[n] = [bill_to_flipper].count()"