This repository contains a minimal starting scaffold for the RezumAI project:
backend/— FastAPI app with placeholder Vertex AI + Agora integrationfrontend/— Next.js app with a simple chat formdocker-compose.yml— brings up both services for local development
Quick run using Docker Compose (PowerShell):
# Build and run services
docker-compose up --buildOpen the frontend at http://localhost:3000 and the backend health at http://localhost:8000/health
Environment variables
- Copy
backend/.env.exampletobackend/.envand fill values for:- VERTEX_PROJECT
- VERTEX_LOCATION
- VERTEX_MODEL_ID
- AGORA_APP_ID
- AGORA_APP_CERTIFICATE
Notes and next steps
- Vertex AI: replace the conservative example in
backend/main.pywith the exact call signature for your model and SDK version. Configure GCP credentials (ADC) in your environment or container. - Agora: replace
generate_agora_tokenwith Agora's official token builder implementation. Add the required server-side SDK as a dependency. - Frontend: currently posts to
http://localhost:8000/chat. If you run behind a proxy or in Docker networking, update the URL to the backend service name or use a relative path with a reverse proxy.
If you'd like, I can now:
- Add a proper Dockerfile for
frontendandbackendif you want them to be self-contained (I noticedDockerfiles already exist in each folder). - Replace the Vertex AI placeholder with a tested code snippet matching a specific Vertex model (tell me which model/resource you plan to use).
- Implement real Agora token generation with the correct Python package (if you provide which Agora product and token type you need).