prepare-benchmark-input-csv

name	prepare-benchmark-input-csv
description	Prepare correct nf-aggregate input CSV files for API runs, external tarballs, external directories, or mixed-source benchmarks. Use when comparing live Seqera runs against exported logs or building reproducible benchmark fixtures.

Prepare benchmark input CSV

Use this skill when the user needs the right --input CSV for nf-aggregate.

Supported row shapes

API rows

For live Seqera Platform runs:

id,workspace
ULVYH4G9KM9sP,seqeralabs/showcase
5HieFbNIsw59ul,seqeralabs/showcase

Grouped API rows

When comparing cohorts:

id,workspace,group
3VcLMAI8wyy0Ld,community/showcase,group1
4VLRs7nuqbAhDy,community/showcase,group2

External tarball rows

For pre-exported run dumps:

id,workspace,group,logs
15dvxY1LnZYrYe,external,g5,logs/15dvxY1LnZYrYe.tar.gz
5ymDk0hmgqv4L4,external,cpu,logs/5ymDk0hmgqv4L4.tar.gz

Mixed live + external rows

id,workspace,group,logs
3VcLMAI8wyy0Ld,community/showcase,live,
1JI5B1avuj3o58,external,external,/path/to/run_dump.tar.gz

Column meanings

• id — Seqera run ID or a stable identifier for the external row
• workspace — org/workspace for live API rows, external for tarball/dir rows
• group — optional comparison group label used in report output
• logs — required for external rows; path to tarball or extracted directory
• platform — optional per-row API endpoint override
• token_env — optional per-row bearer-token env var name

Decision rules

• Use workspace=external only for rows backed by local log artifacts.
• Use logs only for external rows.
• Use group whenever the user wants side-by-side comparisons in the report.
• Keep live and external rows in the same CSV when comparing different environments.

Repo-specific notes

• Test fixtures live under workflows/nf_aggregate/assets/.
• test_benchmark.csv demonstrates tarball-backed benchmark rows.
• test_run_ids.csv demonstrates API-backed rows.
• Mixed-source support is a first-class workflow pattern in this repo.

Validation checklist

Before running the pipeline:

• every live row has a valid org/workspace
• every external row has a readable logs path
• group labels are intentional and consistent
• --generate_benchmark_report is enabled if live API rows should produce report output