data
Health
local dataset inspection · structural health → information density → diversity → novelty → learnability → economic value
Public Atlas →
Analyze a dataset
Compare two
Build reference index
Dataset path (.jsonl / .json / .csv / .tsv / .txt / folder)
Second dataset (B)
Save index to
Options
Text field (auto)
Limit docs
Topics (k)
Near-dup Jaccard
Semantic dedup/diversity (embeddings)
.txt: one doc per line
Force regex NER
Layer 5 (learnability)
Layer 6 (economic)
Reference corpora for novelty — one path per line (.dhidx or dataset)
Layer 5 model (real signals)
Model sample
Fact-density LLM (optional)
Domain priors (JSON, Layer 6)
Run
Point at a dataset and hit Run.