Configuration

Backend environment variables

All backend configuration is loaded from environment variables (or a .env file in the backend directory).

Variable	Required	Default	Description
DATABASE_URL	No	postgres://postgres:postgres@localhost:54322/postgres	PostgreSQL connection string
SUPABASE_JWT_SECRET	Yes	—	JWT secret from Supabase (used to verify auth tokens)
BASE_PATH	No	(empty)	URL prefix for all routes. Empty serves at the domain root; set it when the app lives under a path, as on the CiTIUS instance (/chatty)
HOST	No	0.0.0.0	Address the server binds to
PORT	No	5000	Port the server listens on
OPENAI_KEY	No	—	OpenAI API key (enables GPT-4, GPT-4 Turbo, GPT-3.5 Turbo)
MISTRAL_KEY	No	—	Mistral API key (enables Mistral 7B, Large, Small, Codestral)
GEMINI_KEY	No	—	Google Gemini API key (enables Gemini 1.5 Flash/Pro, 1.0 Pro)
HF_KEY	No	—	HuggingFace API key (enables Flan-T5 Small/Base)
OLLAMA_HOST	No	http://localhost:11434	Ollama server URL (auto-discovers available models)

Note

SUPABASE_JWT_SECRET is the only strictly required variable. You can find it in the output of supabase start or in your Supabase project settings under Settings > API > JWT Secret.

Frontend environment variables

These variables are set in the frontend .env file and must be prefixed with PUBLIC_ (SvelteKit convention).

Variable	Required	Default	Description
PUBLIC_BASE_PATH	No	(empty)	Base path for the SvelteKit app (must match backend BASE_PATH)
PUBLIC_SUPABASE_URL	Yes	—	Supabase API URL (e.g., http://localhost:54321)
PUBLIC_SUPABASE_ANON_KEY	Yes	—	Supabase anonymous key for client-side auth
ORIGIN	No	—	The origin URL, required in production for CSRF protection

API keys and model registration

The backend uses a plugin registry that automatically registers models when the corresponding API key is provided. No key means that provider is skipped silently.

Models registered per provider

Provider	Env var	Models registered
OpenAI	OPENAI_KEY	GPT-4, GPT-4 Turbo, GPT-3.5 Turbo
Mistral	MISTRAL_KEY	Mistral 7B, Mistral Large, Mistral Small, Codestral
Gemini	GEMINI_KEY	Gemini 1.5 Flash, Gemini 1.5 Pro, Gemini 1.0 Pro
HuggingFace	HF_KEY	Flan-T5 Small, Flan-T5 Base
Ollama	OLLAMA_HOST	Auto-discovered from running Ollama instance

Tip

Ollama models are discovered at startup by querying GET {OLLAMA_HOST}/api/tags. Pull any model with ollama pull <model> and restart the backend to pick it up.

Google OAuth setup

Chatty the Lab uses Supabase Auth with Google as the OAuth provider.

1. Create a Google Cloud OAuth client

1. Go to the Google Cloud Console
2. Create a new project or select an existing one
3. Navigate to APIs & Services > Credentials
4. Click Create Credentials > OAuth client ID
5. Select Web application
6. Add authorized redirect URIs:
   • Local development: http://localhost:54321/auth/v1/callback
   • Production: https://your-domain.com/auth/callback (add the base path if you set one)
7. Note the Client ID and Client Secret

2. Configure Supabase Auth

For local development (supabase/config.toml):

[auth.external.google]
enabled = true
client_id = "your-google-client-id"
secret = "your-google-client-secret"
redirect_uri = "http://localhost:54321/auth/v1/callback"

For production (via environment variables in docker-compose.prod.yml):

supabase-auth:
  environment:
    GOTRUE_EXTERNAL_GOOGLE_ENABLED: "true"
    GOTRUE_EXTERNAL_GOOGLE_CLIENT_ID: ${GOOGLE_CLIENT_ID}
    GOTRUE_EXTERNAL_GOOGLE_SECRET: ${GOOGLE_CLIENT_SECRET}
    GOTRUE_EXTERNAL_GOOGLE_REDIRECT_URI: https://tec.citius.usc.es/chatty/auth/callback

3. Verify the configuration

After configuring, restart Supabase (supabase stop && supabase start) and navigate to the login page. Clicking “Sign in with Google” should redirect to the Google consent screen.

Caution

Make sure the redirect URI in Google Cloud Console exactly matches what Supabase Auth expects. A mismatch will cause the OAuth flow to fail with a redirect error.