# Data Curation Tool Modern

A local-first application for curating, tagging, downloading and managing datasets and models, with built-in AI assistants for tag editing, code and dataset-pipeline work, plus 3D generation and media metadata handling.

Data Curation Tool Modern is a local-first application for dataset curation: tagging images and video, auto-tagging, downloading from image boards, managing models, generating 3D assets, and handling media metadata, with a tool-orchestration layer tying these together. Recent updates add app-wide assistant overlays that show live action notes and a visible reasoning trace from any tab while an assistant is running.

It adds system-RAM guardrails for long-running LLM and vision-model sessions, disabling automatic CPU offload by default so a sharded model cannot silently consume all system memory, and bounds stored chat context and response payloads with database-level truncation guards. A tag-editing assistant applies keep, remove, add and final-tag directives to a project's real media tag list based on active dataset-pipeline and LoRA rules. It is written in Python and GPL-3.0 licensed.

Website: https://github.com/x-CK-x/Dataset-Curation-Tool