CLI defaults matter for interpreting a sample
The Node CLI defaults to train and five rows. Its --json output preserves structured answers; terminal percentages are presentation rounding, not additional measurement.
Inspect the supporting source ↗REPOSITORY ANALYSIS
hfjev
Python, Node and playground interfaces for classifying dataset rows with declared Choice, Score and Noul dimensions and exporting structured results.
E2 · source reviewedPublic material or source evidence was reviewed. A fixed-version source review requires checking a pinned revision; this is not a reproduction, performance endorsement, or safety guarantee.
- Stars
- 1
- Forks
- 0
- License
- MIT
- Last commit
- 2026-09-21
What Jev does here
hfjev classifies dataset rows in parallel across typed Choice, Score, and Noul dimensions against TypeSafe System One.
Makes row-level judgments and chosen rubrics inspectable before batch use; playground mock mode is separate from live evaluation.
Fixed-version documentation and implementation material were inspected on 2026-10-24. No dataset rows or live API calls were evaluated. The rubric’s probabilities were not calibrated on a site-owned labeled dataset.
What task does this project fit?
Classify dataset rows across a chosen set of typed dimensions.
Material analysed：
From input to action
- Load a Hugging Face dataset or local file with a configured split and row limit.
- Adapt a dimension pack or supply Choice, Score and Noul dimensions.
- Classify rows and export answers as JSON or human-readable output.
Implementation details that matter
Playground mode changes what the result proves
The key dialog distinguishes direct TypeSafe requests from a client-side mock path. A zero-credit mock demonstration must not be reported as live model performance.
Inspect the supporting source ↗When to choose it
Useful for inspecting rubrics on a small labeled sample before a larger dataset job. Inspect the chosen input column and dimension meanings rather than assuming automatic adaptation is correct.
How to inspect it yourself
Read parseArgs and the result-rendering branches in js/bin/hfjev.js, then inspect ApiKeyModal.tsx’s modes. Confirm limit/split and raw JSON before comparing displayed confidence values.
Limits of this review
- No dataset rows or live API calls were evaluated.
- The rubric’s probabilities were not calibrated on a site-owned labeled dataset.