Match source names to target names using strict name similarity.

The endpoint is a pure matching utility. Callers provide both the source names and target names, and can optionally provide string replacement rules and prefix removal options to apply before matching.

Names are normalized before matching by lowercasing, replacing underscores, dots, dashes, and slashes with spaces, collapsing whitespace, and applying optional replacement rules. Replacement rules match whole normalized token sequences only, so a rule like Fm -> Formation changes Tarbert Fm to Tarbert Formation, while Top -> "" leaves Stop Viking unchanged.

Callers can provide a list of individual prefixes to remove. Prefixes are removed only from the text before the first digit and match normalized tokens. Prefix removal changes only the values used for comparison. The original names remain in the response.

The response contains one result per source. Each result contains up to three target matches, ordered from highest to lowest score.

Confidence levels:

• 'high': score > 80
• 'medium': score 50-80
• 'low': score < 50