feat(meds): OMOP → MEDS conversion for ML-native EHR encoders

.github/workflows/test.yml

@@ -0,0 +1,36 @@
+name: tests
+
+on:
+  pull_request:
+    branches: [main]
+  push:
+    branches: [main]
+
+jobs:
+  pytest:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.11"]
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: pip
+
+      - name: Install package + test extras
+        run: |
+          python -m pip install --upgrade pip
+          pip install -e ".[tests,meds,aws,genomics]"
+
+      - name: Show meds_etl info
+        run: |
+          python -c "import meds_etl; print('meds_etl', meds_etl.__version__)"
+          which meds_etl_omop || true
+
+      - name: Run pytest
+        run: pytest tests/ -v --tb=short

README.md

@@ -40,6 +40,25 @@ SynthLab provides Python interfaces for working with major synthetic healthcare
 - **Access Information**: Clear documentation of open vs. registration-required datasets
 - **Download Utilities**: Helpers for downloading select open-access datasets
 
+### MEDS (Medical Event Data Standard) conversion (NEW)
+Plug synthetic EHR into ML-native foundation models. Provides a thin,
+well-tested wrapper over the community [`meds_etl`](https://github.com/Medical-Event-Data-Standard/meds_etl)
+package so OMOP CSVs (produced by `synthlab.synthea.convert_synthea_to_omop`)
+become MEDS parquet shards ready for models like
+[SMB-v1](https://huggingface.co/standardmodelbio) or MOTOR:
+
+```python
+from synthlab.meds import MedsConvertConfig, convert_omop_to_meds, load_meds_events
+
+convert_omop_to_meds(MedsConvertConfig(
+    omop_dir="~/.cache/synthlab/synthea/omop_100",
+    meds_dir="~/.cache/synthlab/meds/synthea_100",
+))
+df = load_meds_events("~/.cache/synthlab/meds/synthea_100")
+```
+
+Install the optional extra: `pip install synthlab[meds]`.
+
 ## Installation
 
 ```bash

pyproject.toml

@@ -67,6 +67,16 @@ visualization = [
     "pyvis>=0.3.0",
     "networkx>=2.8.0",
 ]
+meds = [
+    # Medical Event Data Standard ETL (OMOP / MIMIC-IV → MEDS parquet).
+    # https://github.com/Medical-Event-Data-Standard/meds_etl
+    "meds_etl>=0.3",
+    "meds>=0.4",
+]
+tests = [
+    "pytest>=8.0",
+    "pytest-cov>=5.0",
+]
 all = [
     "boto3>=1.26.0",
     "numpy>=1.20.0",
@@ -78,6 +88,8 @@ all = [
     "faiss-gpu>=1.7.0",
     "pyvis>=0.3.0",
     "networkx>=2.8.0",
+    "meds_etl>=0.3",
+    "meds>=0.4",
 ]
 
 [project.urls]

synthlab/__init__.py

@@ -145,6 +145,16 @@
     get_snomed_data_dir,
 )
 
+# MEDS (Medical Event Data Standard) conversion for ML-native EHR inputs
+from synthlab.meds import (
+    MedsConvertConfig,
+    convert_omop_to_meds,
+    load_meds_events,
+    get_meds_cache_dir,
+    get_meds_info,
+    print_meds_info,
+)
+
 from synthlab.coherent import (
     list_coherent_components,
     list_coherent_files,
@@ -340,4 +350,11 @@ def _print_banner():
     "SAPBERT_MODEL_ID",
     "SAPBERT_MODELS",
     "EMBEDDING_MODELS",
+    # MEDS conversion
+    "MedsConvertConfig",
+    "convert_omop_to_meds",
+    "load_meds_events",
+    "get_meds_cache_dir",
+    "get_meds_info",
+    "print_meds_info",
 ]