CSV Number Validation — Original Python Demo

Original Python CSV validation demo I built this small example to show how I handle ambiguous numeric inputs without silently changing data. This uses synthetic data and is not a client project. What it checks - IDs stay as text, so 001 retains its leading zeros. - Zero is valid; an empty amount is missing. - Text, NaN and decimal-comma strings are flagged as invalid. - The mean uses only validated amounts and Decimal arithmetic. - No valid amounts produces None, rather than a misleading zero. Verified example 001 / 0 -> valid 002 / blank -> missing 003 / ten -> invalid 004 / 10 -> valid 005 / 1,25 -> invalid 006 / NaN -> invalid Validated mean: 5. Scope This demo validates a numeric column. It does not prove a production integration or a completed customer delivery. Duplicate removal, column mapping and acceptance rules are agreed separately for a paid task. Source and run instructions are available on request.

Companion QA replay For a runnable example of checks that detect regressions in zero, blank and identifier handling, see my original synthetic fixture replay: https://github.com/thefakearchitectstudio-oss/jia-synthetic-qa-fixtures This separate Node.js/Promptfoo example replays local synthetic outputs; it is not a live model benchmark or client case. A fresh run from the public files produced 3 conforming passes, 3 deliberate regression failures and 0 execution errors, with no model API calls.