DEPLOYMENT.md

🦠 BacSense v2 - Hugging Face Deployment Guide

Current Status

• Frontend: Deployed on Vercel ✅
• Backend: Deployed on Hugging Face Spaces (Docker) ⚠️

Issue Diagnosis

The "Not Found" error on Hugging Face typically means:

1. The server isn't starting properly
2. The port configuration is incorrect
3. Model files aren't being found during initialization

Solution Implemented

1. Updated Dockerfile

• Added better debugging with startup script
• Configured proper environment variables for Hugging Face
• Added curl for health checks

2. Added Startup Script (start.sh)

• Shows directory contents on startup
• Verifies model files are present
• Starts uvicorn server with proper configuration

3. API Configuration

• Root endpoint (/) now shows a welcome page
• Health check available at /health
• Debug endpoint at /debug_model to diagnose issues
• API documentation at /docs

How to Deploy

Step 1: Push to Hugging Face

cd "e:\1-Antigravity Python\Bacsense 2.0\BacSense-API"
git add .
git commit -m "Fix deployment: add startup script and debugging"
git push origin main

Step 2: Check Deployment Logs

1. Go to your Hugging Face Space: https://huggingface.co/spaces/joytheslothh/BacSense-API
2. Click on "Logs" tab
3. Watch the startup logs to see if:
   • All dependencies install successfully
   • Model files are found
   • Server starts on port 7860

Step 3: Test Endpoints

Once deployed, test these endpoints:

Health Check:

https://joytheslothh-bacsense-api.hf.space/health

API Documentation:

https://joytheslothh-bacsense-api.hf.space/docs

Debug Model:

https://joytheslothh-bacsense-api.hf.space/debug_model

Frontend Configuration (Vercel)

Update your frontend API URL in Vercel to point to:

https://joytheslothh-bacsense-api.hf.space

Or in your frontend code, update the API base URL:

// In your React code
const API_URL = "https://joytheslothh-bacsense-api.hf.space";

Troubleshooting

If you still see "Not Found":

1. Check the logs - Look for import errors or missing files
2. Verify model loading - Visit /debug_model endpoint
3. Test locally - Run docker build -t bacsense . and docker run -p 7860:7860 bacsense

Common Issues:

TensorFlow Import Error:

• Ensure all requirements are installed
• Check if TensorFlow CPU version is compatible

Model Files Not Found:

• Verify bacsense_v2_package folder exists in the repo
• Check that .keras and .pkl files are tracked by Git LFS

Port Binding Issue:

• Hugging Face expects port 7860
• Server must bind to 0.0.0.0 not localhost

Alternative: Streamlit-Only Deployment

If FastAPI continues to have issues, you can deploy as a pure Streamlit app:

1. Rename app.py to streamlit_app.py
2. Update Dockerfile CMD to: CMD ["streamlit", "run", "streamlit_app.py"]

This gives you the dashboard without the separate API backend.