Deployment and Project Integration · Lesson 4
Deploying GenAI Apps via Hugging Face Spaces and Intro to GCP
Hugging Face Spaces hosts demos for free or cheap using Gradio, Streamlit or a custom Docker image. Push a repo with your app file and requirements, add API keys as Space secrets, and it builds automatically.
Google Cloud Platform (GCP) offers production hosting. Cloud Run runs containers serverlessly and scales to zero; Artifact Registry stores images; Secret Manager stores keys; Vertex AI provides managed models.
Typical path: prototype on Spaces, then containerise and deploy to Cloud Run for production traffic.
Key points
- Spaces: quick demos with Gradio/Streamlit/Docker
- Store keys as secrets, never in code
- Cloud Run runs containers serverlessly
- Vertex AI offers managed models on GCP