Initial release: rarelens platform skeleton (AGPL-3.0)

End-to-end variant interpretation platform for rare genetic disease research:
SvelteKit UI, FastAPI + PostgreSQL API, Nextflow/Ensembl VEP pipeline,
LightGBM pathogenicity scoring with MLflow, K8s/ArgoCD/GCP infrastructure.
Public test data only; no clinical claims.

api/app/services/scoring.py

@@ -0,0 +1,45 @@
+import uuid
+
+import mlflow
+import pandas as pd
+from sqlalchemy import select
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.config import settings
+from app.models import Prediction, Variant
+
+_model = None
+
+
+def load_model():
+    global _model
+    if _model is None:
+        mlflow.set_tracking_uri(settings.mlflow_tracking_uri)
+        _model = mlflow.pyfunc.load_model(f"models:/{settings.model_name}/{settings.model_stage}")
+    return _model
+
+
+def featurise(variants: list[Variant]) -> pd.DataFrame:
+    # Mirror ml/rarelens_ml/features.py exactly; shared package later.
+    return pd.DataFrame(
+        {
+            "impact": [v.impact for v in variants],
+            "consequence": [v.consequence for v in variants],
+            "gnomad_af": [v.gnomad_af if v.gnomad_af is not None else 0.0 for v in variants],
+            "cadd_phred": [v.annotations.get("CADD_PHRED") for v in variants],
+            "am_pathogenicity": [v.annotations.get("am_pathogenicity") for v in variants],
+        }
+    )
+
+
+async def score_job(job_id: uuid.UUID, session: AsyncSession) -> int:
+    variants = (await session.scalars(select(Variant).where(Variant.job_id == job_id))).all()
+    if not variants:
+        return 0
+    model = load_model()
+    scores = model.predict(featurise(variants))
+    for v, s in zip(variants, scores):
+        session.add(Prediction(variant_id=v.id, model_name=settings.model_name,
+                               model_version=settings.model_stage, score=float(s)))
+    await session.commit()
+    return len(variants)