feat: add uvicorn fallback for server startup and install runpod into existing virtual environment

Author: arkodeepsen

Dockerfile.wrapper

@@ -1,15 +1,13 @@
 # Use the existing working Kokoro FastAPI image as base
 FROM ghcr.io/remsky/kokoro-fastapi-gpu:latest
 
-# Switch to root to install additional packages
-USER root
-
-# Install runpod and any additional dependencies
-RUN pip install runpod>=1.0.0
-
 # Copy the serverless handler
 COPY handler-wrapper.py /app/handler.py
 
+# Switch to root to install runpod in the existing virtual environment
+USER root
+RUN /app/.venv/bin/pip install runpod>=1.0.0
+
 # Switch back to appuser
 USER appuser
 

handler-wrapper.py

@@ -26,10 +26,20 @@ def start_fastapi():
 
     logger.info("Starting internal FastAPI server...")
 
-    # Start the FastAPI server using the existing entrypoint
-    fastapi_process = subprocess.Popen([
-        "/app/entrypoint.sh"
-    ], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
+    # Start the FastAPI server using the existing startup method
+    # Try entrypoint.sh first, then fallback to direct uvicorn
+    try:
+        fastapi_process = subprocess.Popen([
+            "/app/entrypoint.sh"
+        ], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
+    except FileNotFoundError:
+        # Fallback to direct uvicorn start if entrypoint.sh doesn't exist
+        fastapi_process = subprocess.Popen([
+            "/app/.venv/bin/python", "-m", "uvicorn",
+            "api.src.main:app",
+            "--host", "0.0.0.0",
+            "--port", "8880"
+        ], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True, cwd="/app")
 
     # Wait for the server to be ready
     max_wait = 120  # 2 minutes max wait