Dataset quality score: a free ADC preview before you publish
Paste a sample of your dataset (CSV, TSV, JSON, or JSONL). The analyzer scores it on completeness, schema consistency, and type consistency (three signals from Sella’s Application Data Contract) and flags issues you’d hit at publish time. Runs in your browser, no upload.
Ready to publish? Take this sample and the full file to /dashboard/publish where Sella runs the full ADC evaluation, generates a datacard, and lists it on the marketplace.
What the three score dimensions mean
Completeness measures how often each column has a non-empty value. A score below 90% on a primary column is usually a publish-blocker. Agents expect to receive a value. Schema consistency checks that every row has the expected columns or fields. Drift here breaks deterministic downstream parsers. Type consistency looks at whether each column holds a single inferred type (number, string, date). Mixed-type columns are a red flag for typed agent clients.
Once you're happy with the score, publish on Sella and the platform runs the full ADC evaluation, generates a datacard, and lists the dataset on the agent marketplace.
FAQ
- What is an ADC score?
- Application Data Contract: the structural quality bar Sella uses to evaluate datasets at publish time. The full evaluator runs deeper checks (PII detection, modality match, drift), but the three signals here (completeness, schema consistency, type consistency) are the load-bearing ones.
- How big a sample should I paste?
- Aim for 50–500 rows. Smaller samples are still informative but the analyzer flags low-confidence scores below 10 rows. The full file gets a deeper evaluation when you publish on Sella.
- Does my data leave my browser?
- No. The analyzer runs entirely in your browser. Nothing is uploaded.
Sella, the economy for AI agents
Discover, preview, and pay for machine-payable datasets, evals, and APIs on the Sella marketplace. Or publish your own.