Skip to content

Feedback wanted: Promptfoo JSONL imports in v0.2.25 #50

Description

@CAOShurong

EvalInt v0.2.25 can now read Promptfoo's current per-result JSONL export through automatic detection or --format promptfoo.

Release: https://github.com/CAOShurong/evalint/releases/tag/v0.2.25
Contract and research: https://github.com/CAOShurong/evalint/blob/v0.2.25/docs/PROMPTFOO_JSONL.md

This is a maintainer-created feedback request, not evidence of external users, adoption, or broad compatibility.

Useful feedback would include:

  • the exact Promptfoo version and command that produced the artifact;
  • whether automatic detection selected promptfoo;
  • whether item/provider counts match the producer;
  • any PROMPTFOO_STRIP_* options used;
  • a minimal synthetic or carefully sanitized row when a current export is rejected;
  • whether unrelated JSONL was falsely classified as Promptfoo.

Please do not upload secrets, real prompts, model outputs, provider configuration, or private eval data. Reduce a report to a synthetic minimal fixture first.

Known boundaries: EvalInt still reads the complete file text into memory; JSONL support is not a streaming or denial-of-service guarantee. If variables and prompt text are stripped, testIdx is only an export-local identity and should not be assumed stable across runs. A clean audit does not prove that all expected results were exported, that the artifact is authentic, or that its scores are semantically correct.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions