Skip to content

data ingest: data-first flow + task picker + --task/--split/--name (RFC-0002 phase 2) #180

Description

@LukasWodka

Phase 2 of RFC-0002 — the core UX change. Invert the flow to data-first and rework task selection.

  • Order: train/test → name → point at your data → task.
  • Sniff the family from the data + echo it ("✔ Found a CSV table — tabular data"); show only the compatible tasks, each as Name — one-liner · id with a guiding example; split Available now vs CLI-pending.
  • Rename --category--task (required, hidden alias, drop the image default), --intent--split (default train), --table--name.
  • Label = pick from the real CSV headers (exact-match on the flag path), which kills the data-ingestors#340 case-mismatch silent-null-label bug; per-task wording (class vs value-to-predict); skipped for self-supervised tasks.

Design-of-record: RFC-0002 (#174). Part of the data-ingest redesign epic. (§5, §7, §8)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions