Trust and operation
Methodology
UChartX uses transparent browser-side rules for parsing, chart recommendation, audit scoring, and export readiness.
Last updated: May 12, 2026Key points
- Field types are inferred from visible sample values.
- Chart recommendations are based on data shape and stated task.
- Audit scores combine labels, axes, data, color, readability, and source context.
Parsing and field inference
CSV, JSON, Markdown, Excel, and pasted rows are normalized into a table. Field types are inferred from the visible values.
- Numeric fields are detected after common symbols are removed.
- Dates are detected from parseable date-like values.
- Nested JSON objects are flattened into dot-separated field names.
Recommendation and scoring
Recommendations compare the field mix with the chart task. The score uses six checks that add up to 100 points before data and shape penalties.
- Data readiness 20, field mapping 15, chart fit 20.
- Labels and context 15, accessibility and color 15, export readiness 15.
- The listed fixes explain what changed the score and what to review next.