Semantic search (embedding-based)
/construe/codes/{codesystem}/search/semanticPerforms semantic similarity search using vector embeddings.
Availability: This endpoint works for both built-in and custom code systems.
When to use: Best for natural language queries where you want to find conceptually related codes, even when different terminology is used. The search understands meaning, not just keywords.
Examples:
- Query "trouble breathing at night" finds codes like "Sleep apnea", "Orthopnea", "Nocturnal dyspnea" — semantically related but no exact keyword matches
- Query "heart problems" finds "Myocardial infarction", "Cardiac arrest", "Arrhythmia"
Trade-offs: Slower than text search (requires embedding generation), but finds conceptually similar results that keyword search would miss.
See also: /search/text for faster keyword-based lookup with typo tolerance.
Usage of CPT is subject to AMA requirements: see PhenoML Terms of Service.
Path parameters
codesystemstringrequiredCode system name
Query parameters
textstringrequiredNatural language text to find semantically similar codes for
versionstringoptionalSpecific version of the code system, such as umls-2026aa.
limitintegeroptionaldefault 10Maximum number of results (default 10, max 50)
Semantic search results ordered by similarity
Response fields
systemobjectrequirednamestringrequiredCode system name
versionstringrequiredCode system version
resultsobject[]requiredCodes ordered by semantic similarity (most similar first)
codestringrequireddescriptionstringrequired