Find an NUCC taxonomy by specialty
Heuristichealthcare_taxonomy_searchBest-effort — fuzzy, approximate, or inferred; check the confidence score.
Search NUCC taxonomy by free text. Three modes: 'exact' (case-insensitive equality on display_name or code — fast and narrow), 'fuzzy' (rapidfuzz token_set_ratio over display_name — sub-ms, works without the embedding server), 'semantic' (vector KNN via the local embedding service — highest quality). Default mode is 'semantic'; it falls back to 'fuzzy' when the embed server is unreachable (the response's `mode_used` reflects what actually ran). Returns up to `limit` candidates ordered by score (0.0-1.0). Top-level confidence equals the best candidate's score. Processing multiple inputs? Use `healthcare_taxonomy_search_batch` to do up to 200 in one call.
Parameters
querystringrequiredFree-text query, e.g. 'family medicine'.
modestringoptionalSearch mode: 'exact' (case-insensitive equality on display_name or code), 'fuzzy' (rapidfuzz over display_name), or 'semantic' (vector KNN via the embedding service). Semantic mode falls back to fuzzy when the embed server is unavailable.
limitintegeroptionalfieldsarrayoptionalOptional list of fields to return per row. Defaults to all. Valid fields: code, grouping, classification, specialization, definition, notes, display_name, section. The 'code' field is always returned.
Call it
curl -X POST https://api.snipget.ai/v1/healthcare/taxonomy/search \
-H "Authorization: Bearer sk_live_..." \
-H "Content-Type: application/json" \
-d '{"query":"family medicine","mode":"fuzzy"}'The Batch tab sends up to 1000 inputs in one call via /v1/healthcare/taxonomy/search/batch.
FAQ
When should I use search?
Use it when you have a specialty in words, like 'family medicine', but not the NUCC code. It returns ranked taxonomy candidates from the maintained dataset so you can pick the right code.
How exact are the results?
This is a best-effort, ranked search, not an exact lookup. Read the per-result score, which runs from 0 to 1, and choose accordingly rather than assuming the top hit is always right.
What are the search modes?
'exact' for a case-insensitive equality match, 'fuzzy' for approximate text matching that always works, and 'semantic' for meaning-based matching. Semantic falls back to fuzzy if its service is unavailable, and the response tells you which mode actually ran.
What if I already know the exact code?
Use the taxonomy lookup tool instead, which returns the single authoritative row for that code without ranking.