Skip to content

Post-JSONL-split follow-ups (offline readers, per-output split config) #52

Description

@ff225

Priority: Low

Follow-ups after the JSONL segment-splitting change.

Tasks

  • Verify offline notebooks/scripts read all segments via run_segment_paths (or an iter_run_* helper) instead of a single file — e.g. scripts/analysis/*, offloading_decision_summary.py, src/server/evaluations/edit_evaluations.ipynb. With splitting enabled the base file is no longer written.
  • Evaluate a dedicated per-output split config (e.g. output.split) if the two JSONL outputs ever need different policies. Today a single shared knob (evaluation.split) governs both.

From the Feature 042 codebase review (recommendations #11, #13).

Metadata

Metadata

Assignees

No one assigned

    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