HN Showcase
Back to browse
GitHub Repository

Research Archival & Integrity Recorder - a Simple Data Versioning Tool

2 starsPython

Reproducible experiments without committing every tweak

by Saloc·Jun 4, 2026·2 points·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidSolve My ProblemBig Brain

Git diff tracking without commit clutter solves the real experiment iteration pain.

Strengths
  • •Captures uncommitted parameter changes as patches alongside git references
  • •Hash-based deduplication avoids storing duplicate data files automatically
  • •Zero-config heuristics detect inputs and outputs without manual setup
Weaknesses
  • •Experiment tracking already served by DVC, MLflow, and Weights & Biases
  • •Niche audience limits broader developer adoption
Target Audience

Researchers and ML engineers running iterative experiments

Similar To

DVC · MLflow · Sacred

Post Description

I hate keeping extensive notes of experiments. When iterating fast, it feels to break the flow. However when keeping all adjustments in mind - its only a question of time until valuable info gets lost.

Therefore I build Rair to track results, inputs and code with minimal overhead. Putting "rair" in front of the command without any config is often sufficient.

For code tracking, Rair references git, but it also tracks uncommitted changes as diff - like adjustments of parameters in source files.

Input and result data is detected heuristically if not configured. Data tracking is based on file hashes for deduplication.

I'm curious whether this "just prefix the command" approach feels useful for others.

Similar Projects