"Target column has too few unique values" — the chosen target barely varies. Pick a target with real spread, or mark it categorical if it is a category label.
"Column types look wrong" — the auto-detector mislabeled a column. Review the detected types and correct them before rerunning.
"Dataset exceeds tier limits" — reduce rows/columns, or upgrade the tier.
Timeout / memory errors — usually generated by very large synthetic batches or heavy feature engineering. Reduce rows per run or drop low-value features.
Upload problems
File over the size limit → the error states the limit.
Duplicate column headers → rename columns and retry.
Mixed types in one column (numbers + text) → clean the column; the detector will otherwise classify it as categorical.
Unexpected results
Weak evaluation scores → try a different data synthesis mode (Statistical on small data), generate fewer rows, or fix column types.
Many extrapolation flags → the model is being asked to predict outside your data's envelope. Extend the input ranges you trained on, or treat flagged predictions as hypotheses.
Still stuck?
The AI Copilot answers with citations into these docs. For real bugs, use the Contact page's "Bugs / Issues" category and include the job ID.