Local-first dataset workbench for visual QA, human-approved releases, and traceable metrics.
-
Updated
Nov 17, 2026 - Python
Local-first dataset workbench for visual QA, human-approved releases, and traceable metrics.
This is the public code repository of the BEXIS2 data management software. It contains only modules, components, and packages of the core system. Contributed modules and components will be available in separate repositories. For more information on BEXIS2, please visit our website.
Local content-addressed archive with observation history. Stores bytes by SHA-256, preserves locator state as contiguous spans, compresses with zstd and corpus-trained dictionaries. SQLite-backed.
Python Library that mantains ML pipelines with data checks, plots, training, and versioning. Check the clear documentation here:
Local-first pipeline to turn business feedback and corrections into specialized MLX models on Apple Silicon with traceable datasets, LoRA fine-tuning, and regression gating.
Data Version Control (DVC) learning repository for Machine Learning Engineering, covering dataset versioning, experiment tracking, pipeline management, reproducible ML workflows, remote storage integration, and collaborative model development.
FaultLine is a cloud-native geospatial change engine for comparing, validating, and tracking spatial dataset evolution across modern formats like GeoParquet, COG, and PMTiles.
Top Dataset Versioning Platform (Opensource) 🌟 Star if you like it! 🌟
CairnML SDK: unofficial modified fork of the ClearML SDK
Dataset versioning platform inspired by modern ML infrastructure systems like Hugging Face Hub, built on Apache Iceberg, Spark, and S3.
Auditable annotation workbench for CSV and JSONL: bounded imports, duplicate quarantine, immutable schema versions, revision-checked autosave, review and adjudication, honest agreement metrics, and SHA-256 dataset releases. Next.js 16, TypeScript, PostgreSQL, Prisma. Synthetic live demo with owner, reviewer, and annotator roles. MIT.
Lightweight ML experimentation toolkit: versioned datasets, reproducible runs, experiment tracking, cloud GPUs. No servers.
A reproducible MLOps experiment for publishing and validating tamper-evident datasets.
One-command determinism + manifest for ML projects.
A lockfile for your ML datasets and experiments — reproduce any run without copying a single frame
Test the data behind an AI system like code: eval contamination, wrong reference labels, unrepresentative eval sets, stale and contradictory RAG documents, personal data, data contracts and versioning, each gated in CI. No API key needed.
Mirror models, datasets, mcps, and spaces from GitHub to MEGA Hub
To associate your repository with the dataset-versioning topic, visit your repo's landing page and select "manage topics."