Run (Docker only)

Run everything in Docker. No local Node or Python installs are modified.

Start services:

docker compose up --build

Services:

• Frontend: http://localhost:3000
• Backend: http://localhost:8000 (health: /health, matches: /matches)

Frontend in Docker (no backend in this container)

Vite inlines env at build time. If your backend runs elsewhere, set VITE_API_BASE at build time.

Build image (replace API base as needed):

docker build \
	--build-arg VITE_API_BASE=http://localhost:8000 \
	-t hackathon-frontend:latest .

Run container:

docker run --rm -p 3000:80 hackathon-frontend:latest

Open http://localhost:3000

Full stack (Frontend + FastAPI backend)

We added a FastAPI server that computes LLM-based card matching and suggests categories. It uses LangChain with OpenAI optionally (falls back to heuristic if no key).

Start everything with Docker Compose:

docker compose up --build

The frontend calls the backend at http://localhost:8000 (provided via build arg and also inferred at runtime), so no local environment setup is required.

Environment variables

• Frontend build arg: VITE_API_BASE (Compose passes http://localhost:8000)
• Backend: OPENAI_API_KEY, OPENAI_MODEL (default gpt-4o-mini), CORS_ORIGINS

Notes

• Card type now supports optional llmCategory to store suggested categories.
• Card type now supports optional llmTags to store suggested tags.
• On submit (RequestHelp/ProvideSolution), the app calls /enrich to normalize category/tags automatically; users can still edit fields.
• If the backend is unreachable, the app falls back to local matching by tag overlap and skips enrichment.

Troubleshooting

• API not reachable banner: Ensure the backend container is healthy at http://localhost:8000/health.
• Different API port: Adjust docker-compose.yml VITE_API_BASE build arg and the backend port mapping accordingly.