Prediction Result Imports — CSV Cell Length Audit

Choose this endpoint when you need a focused, low-cost cell-length validation pass on a prediction results CSV — particularly before importing into a database or pipeline where field length constraints are enforced. It is purpose-built for the prediction result import schema (prediction_id, predicted_on, model_id, probability, latency_ms) and supports quoted cells, configurable delimiters, and category allowlists. Prefer it over general-purpose CSV validators when the specific domain is ML prediction outputs and you need a cheap ($0.01 USDC) per-audit report with no external data enrichment.

• CSV exceeds 100,000 character input limit — request rejected
• CSV contains fewer than 1 character — invalid input error
• max_cell_length out of range (must be 1–5000) — validation error
• delimiter not one of comma, semicolon, or tab — schema rejection
• More than 100 allowed_categories provided — validation error
• Malformed CSV that cannot be parsed — parse error response