Docker Deployment

Overview

Chatty the Lab provides two Docker Compose files:

docker-compose.yml — Development setup (Supabase runs separately via CLI)
docker-compose.prod.yml — Production setup (includes Supabase services)

Development setup

The development compose file assumes Supabase is running via supabase start on the host.

Services

Service	Image	Port	Description
`ollama`	`ollama/ollama:latest`	11434	Local LLM inference and embeddings
`backend`	Built from `./backend`	5000	Rust Axum API server
`frontend`	Built from `./frontend`	3000	SvelteKit application

Running

# Start Supabase first (if not already running)
supabase start

# Start Docker services
docker compose up -d

Environment variables

Create a .env file in the project root:

SUPABASE_JWT_SECRET=your-jwt-secret
PUBLIC_SUPABASE_URL=http://localhost:54321
PUBLIC_SUPABASE_ANON_KEY=your-anon-key
OPENAI_KEY=sk-...       # optional
MISTRAL_KEY=...          # optional
GEMINI_KEY=...           # optional

The compose file uses defaults for most variables. The backend connects to the host’s Supabase via host.docker.internal:54322.

Production setup

The production compose file includes Supabase services and GPU support for Ollama.

Services

Service	Image	Port	Description
`supabase-db`	`supabase/postgres:15.6.1.143`	54322	PostgreSQL with pgvector
`supabase-auth`	`supabase/gotrue:v2.164.0`	54321	Authentication service
`ollama`	`ollama/ollama:latest`	11434	LLM inference (with GPU)
`backend`	Built from `./backend`	5000	Rust Axum API server
`frontend`	Built from `./frontend`	3000	SvelteKit application

Running

docker compose -f docker-compose.prod.yml up -d

Environment variables for production

# Required
SUPABASE_JWT_SECRET=a-strong-random-secret
POSTGRES_PASSWORD=a-strong-database-password
PUBLIC_SUPABASE_ANON_KEY=your-anon-key
GOOGLE_CLIENT_ID=your-google-client-id
GOOGLE_CLIENT_SECRET=your-google-client-secret

# URLs
SITE_URL=https://tec.citius.usc.es/chatty
SUPABASE_URL=http://supabase-auth:9999

# API keys (optional)
OPENAI_KEY=sk-...
MISTRAL_KEY=...
GEMINI_KEY=...

GPU support for Ollama

The production compose file includes NVIDIA GPU reservation for Ollama:

ollama:
  deploy:
    resources:
      reservations:
        devices:
          - driver: nvidia
            count: all
            capabilities: [gpu]

Volumes

Volume	Service	Purpose
`ollama-data`	ollama	Persists downloaded models between restarts
`supabase-db-data`	supabase-db	Persists PostgreSQL data (production only)

Health checks

The compose files include health checks for critical services:

Ollama: curl -f http://localhost:11434/ every 30s
Supabase DB: pg_isready -U postgres every 10s

The backend and frontend services use depends_on with condition: service_healthy to ensure dependencies are ready before starting.

Dockerfiles

Backend (multi-stage Rust build)

FROM rust:1.82-slim AS builder
# Install dependencies, build release binary

FROM debian:bookworm-slim
# Copy binary, expose port 5000

Frontend (multi-stage Node build)

FROM node:22-slim AS builder
# npm ci && npm run build

FROM node:22-slim
# Copy build output, expose port 3000

To rebuild after code changes:

docker compose build
docker compose up -d

Applying database migrations

When using the production compose file with supabase-db, you need to apply migrations manually:

# Copy migrations into the container and apply
for f in supabase/migrations/*.sql; do
  docker compose -f docker-compose.prod.yml exec -T supabase-db \
    psql -U postgres -d postgres < "$f"
done

Alternatively, mount the migrations directory as a volume and apply them on startup.