Environment Variables

All backend configuration is done via environment variables. The backend reads from a .env file in the working directory or from the system environment.

Core

Variable	Default	Description
`APP_ENV`	`development`	Environment name. Set to `production` in prod.
`DATABASE_URL`	`sqlite:///./xinference_saas.db`	SQLAlchemy database URL
`DATABASE_AUTO_CREATE_SCHEMA`	`false`	Auto-create DB tables on startup (dev only)

Auth & Sessions

Variable	Default	Description
`SESSION_COOKIE_SECURE`	`false`	Require HTTPS for session cookies. Must be `true` in production.
`SESSION_COOKIE_SAMESITE`	`lax`	Cookie SameSite policy: `lax`, `strict`, or `none`
`SESSION_COOKIE_MAX_AGE`	`604800`	Cookie TTL in seconds (default: 7 days)
`SESSION_EXPIRE_MINUTES`	`10080`	Session expiry in minutes (default: 7 days)
`SESSION_COOKIE_NAME`	auto	Override the session cookie name

Warning

SESSION_COOKIE_SAMESITE=none requires SESSION_COOKIE_SECURE=true. The backend validates this combination on startup and will refuse to start with an unsafe configuration.

CORS

Variable	Default	Description
`CORS_ALLOW_ORIGINS`	`` (empty)	Comma-separated list of allowed origins
`XINFERENCE_FRONTEND_URL`	`/`	URL of the frontend. Must be included in `CORS_ALLOW_ORIGINS`.

Same-site example (app.example.com + api.example.com):

CORS_ALLOW_ORIGINS=https://app.example.com
XINFERENCE_FRONTEND_URL=https://app.example.com
SESSION_COOKIE_SECURE=true
SESSION_COOKIE_SAMESITE=lax

Cross-site example (app.example.com + api.other-domain.com):

CORS_ALLOW_ORIGINS=https://app.example.com
XINFERENCE_FRONTEND_URL=https://app.example.com
SESSION_COOKIE_SECURE=true
SESSION_COOKIE_SAMESITE=none

Google SSO

Variable	Description
`GOOGLE_OIDC_CLIENT_ID`	Google OAuth client ID
`GOOGLE_OIDC_CLIENT_SECRET`	Google OAuth client secret
`GOOGLE_OIDC_REDIRECT_URI`	OAuth callback URL (must match Google Console)

AWS

Variable	Default	Description
`XINFERENCE_SAAS_CLOUD_PROVIDER`	`fake`	Set to `aws` for real provisioning
`XINFERENCE_SAAS_AWS_REGION`	—	AWS region (e.g. `us-east-1`)
`XINFERENCE_SAAS_AWS_AMI_ID`	—	AMI ID with Xinference pre-installed
`XINFERENCE_SAAS_AWS_SUBNET_ID`	—	Subnet ID for EC2 instances
`XINFERENCE_SAAS_AWS_SECURITY_GROUP_IDS`	—	Comma-separated security group IDs
`XINFERENCE_SAAS_AWS_IAM_INSTANCE_PROFILE_NAME`	—	IAM instance profile for workers
`XINFERENCE_SAAS_AWS_KEY_NAME`	—	EC2 key pair name (optional, for SSH access)
`XINFERENCE_SAAS_AWS_ASSOCIATE_PUBLIC_IP`	`false`	Assign public IPs (for testing outside VPC)

Xinference Worker

Variable	Default	Description
`XINFERENCE_SAAS_XINFERENCE_CLIENT`	`fake`	Set to `real` for real Xinference calls
`XINFERENCE_SAAS_XINFERENCE_READY_TIMEOUT_SECONDS`	`600`	Max wait for model to load
`XINFERENCE_SAAS_XINFERENCE_TIMEOUT_SECONDS`	`30`	Per-request HTTP timeout
`XINFERENCE_SAAS_XINFERENCE_IMAGE`	`xprobe/xinference:latest`	Docker image for Xinference
`XINFERENCE_SAAS_XINFERENCE_API_KEY`	—	API key for Xinference server

Cluster Behavior

Variable	Default	Description
`XINFERENCE_SAAS_CLUSTER_IDLE_TTL_SECONDS`	`900`	Seconds of inactivity before auto-termination (default: 15 min)
`XINFERENCE_SAAS_SUPERVISOR_ENDPOINT_HOST`	`private`	Use `private` (VPC) or `public` IP for supervisor
`XINFERENCE_SAAS_AWS_SUPERVISOR_INSTANCE_TYPE`	`t3.small`	Instance type for cluster supervisor node
`XINFERENCE_SAAS_RECONCILER_ENABLED`	`false`	Enable background reconciler loop
`XINFERENCE_SAAS_RECONCILER_INTERVAL_SECONDS`	`60`	How often the reconciler runs

Billing & Stripe

Billing is backed by Stripe. Configure the following to enable subscriptions, payment methods, invoices, and webhooks.

Variable	Description
`STRIPE_SECRET_KEY`	Stripe API secret key
`STRIPE_PUBLISHABLE_KEY`	Stripe publishable key (returned by `GET /v1/billing/config`)
`STRIPE_WEBHOOK_SECRET`	Secret used to verify Stripe webhook signatures (`POST /v1/billing/webhook`)
`STRIPE_DEFAULT_COLLECTION_METHOD`	Default subscription collection method (`charge_automatically` or `send_invoice`)
`STRIPE_CHECKOUT_SESSION_CURRENCY`	Currency used for Stripe Checkout sessions
`STRIPE_BILLING_PORTAL_RETURN_PATH`	Return path after the Stripe customer portal
`STRIPE_SUBSCRIPTION_PAYMENT_BEHAVIOR`	Stripe subscription payment behavior
`STRIPE_SUBSCRIPTION_DAYS_UNTIL_DUE`	Days until an invoice is due (for `send_invoice`)
`BILLING_STRIPE_CATALOG_NAMESPACE`	Namespace for the Stripe billing catalog
`BILLING_SEED_DEFAULT_PRICE_PLAN`	Seed the `aws-ondemand-default` price plan on startup only when no active default plan exists

Email

Variable	Description
`SMTP_HOST`	SMTP server hostname
`SMTP_PORT`	SMTP port (typically `587` or `465`)
`SMTP_USERNAME`	SMTP username
`SMTP_PASSWORD`	SMTP password
`SMTP_USE_TLS` / `SMTP_USE_SSL`	Transport security toggles
`EMAIL_FROM`	Sender address for transactional emails
`INVITATION_EMAIL_FROM` / `INVITATION_EMAIL_REPLY_TO`	Sender and reply-to addresses for invitation emails

Leave SMTP_HOST empty to skip email delivery. Password-reset abuse controls (PASSWORD_RESET_EMAIL_COOLDOWN_SECONDS, PASSWORD_RESET_IP_WINDOW_SECONDS, PASSWORD_RESET_IP_MAX_REQUESTS) are also configurable.