Python

Python notebooks and data science pipelines should start with the convenience facade. It keeps the connection and engine setup out of the main workflow while still returning the same public model objects.

import medterm4ds as mt

terms = mt.connect("/mnt/d/medterm4ds/data/umls_current.duckdb")

info = terms.lookup("ICD10CM", "E11.9")
friendly = terms.patient_friendly("ICD10CM", "E11.9")
mappings = terms.map("ICD10CM", "E11.9", target_sources=["SNOMEDCT_US"])

Batch calls use a source plus a list of codes:

codes = ["E11.9", "E11.40", "E11.42"]

lookup_df = terms.lookup_df("ICD10CM", codes)
friendly_df = terms.patient_friendly_df("ICD10CM", codes)
mapping_df = terms.map_df("ICD10CM", codes, target_sources=["SNOMEDCT_US"])

Mixed-source batches should use CodeRef objects:

refs = [
    mt.CodeRef("ICD10CM", "E11.9"),
    mt.CodeRef("CVX", "208"),
    mt.CodeRef("RXNORM", "1049502"),
]

terms.lookup_df(refs)

Use with when a short script should close the DuckDB connection automatically:

with mt.connect("/mnt/d/medterm4ds/data/umls_current.duckdb") as terms:
    df = terms.search_df("metformin", sources=["RXNORM"], limit=20)

The same facade can wrap the remote API engine:

terms = mt.connect_remote("http://localhost:8000")
terms.lookup("ICD10CM", "E11.9")

Advanced users can still use engines and service functions directly when they need lower-level control:

import duckdb
from medterm4ds import CodeRef, get_code_infos
from medterm4ds.engines.duckdb import LocalDuckDBEngine

con = duckdb.connect("/mnt/d/medterm4ds/data/umls_current.duckdb", read_only=True)
engine = LocalDuckDBEngine(con)

rows = get_code_infos([CodeRef("ICD10CM", "E11.9")], engine=engine)