reduce container size

Dockerfile

@@ -1,26 +1,42 @@
-FROM tiangolo/uvicorn-gunicorn:python3.11
-
-COPY ./requirements.txt /tmp/requirements.txt
+# ---- Build stage ----
+FROM python:3.11-slim-bullseye
+
+COPY requirements.txt .
+COPY requirements-dev.txt .
+    
+ARG TEST
+
+# Installer les dépendances
+RUN pip install --upgrade pip && \
+    pip install --no-cache-dir -r requirements.txt && \
+    if [ "$TEST" = "true" ]; then \
+      pip install --no-cache-dir -r requirements-dev.txt; \
+    fi
+
+# Installer curl pour le healthcheck
+RUN apt-get update && apt-get install -y --no-install-recommends curl && \
+    rm -rf /var/lib/apt/lists/*
+
+# Copier le code
 COPY ./entrypoint.sh /tmp/entrypoint.sh
 COPY ./src /app/src
 COPY ./tests /app/tests
 
-RUN pip install --no-cache-dir -r /tmp/requirements.txt
-
 WORKDIR /app
 
+# Pour que les imports soient résolus depuis /app
 ENV PYTHONPATH=/app
 
-# Port to expose
+# Exposer le port
 EXPOSE 7860
 
-# Health Check
+# Healthcheck sur FastAPI
 HEALTHCHECK --interval=30s --timeout=30s --start-period=5s --retries=3 \
-  CMD [ "curl", "-f", "http://localhost:7860/check_health" ]
+  CMD curl -f http://localhost:7860/check_health || exit 1
 
-# Create a non-root user 'appuser' and switch to this user
+# Utilisateur non-root pour la sécurité
 RUN useradd --create-home appuser
 USER appuser
 
-# CMD with JSON notation
-CMD ["/tmp/entrypoint.sh"]
+# Entrypoint (par exemple pour lancer uvicorn)
+CMD ["/tmp/entrypoint.sh"]

requirements.txt

@@ -1,10 +1,11 @@
-python-dotenv
-fastapi
-psycopg2-binary
-pandas
-scikit-learn
-openpyxl
-xlrd >= 2.0.1
-mlflow
-boto3
-pytest
+python-dotenv==1.1.0
+fastapi==0.115.12
+psycopg2-binary==2.9.10
+pandas==2.2.3
+scikit-learn==1.6.1
+openpyxl==3.1.5
+xlrd==2.0.1
+mlflow-skinny==2.21.3
+boto3==1.37.34
+uvicorn==0.34.1
+gunicorn==23.0.0

src/model.py

@@ -276,10 +276,17 @@ def list_registered_models() -> List[Dict]:
     List all the registered models
     """
     # Set tracking URI to your Heroku application
-    mlflow.set_tracking_uri(os.environ["MLFLOW_SERVER_URI"])
-    
-    # Return the list of registered models
-    results = mlflow.search_registered_models()
+    tracking_uri = os.environ.get("MLFLOW_SERVER_URI")
+    if tracking_uri is None:
+        raise ValueError("MLFLOW_SERVER_URI environment variable is not set.")
+    print(f"MLflow tracking URI: {tracking_uri}")
+
+    client = MlflowClient(tracking_uri=tracking_uri)
+    # Should be:
+    #   results = client.search_registered_models()
+    # but this is not working from inside the container
+    # so we need to use the store client to get the registered models
+    results = client._get_registry_client().store.search_registered_models()
 
     output = []
     for res in results: