Environment & Configuration

All Svantic deployment configuration is managed through environment variables. No code changes are required to switch between topologies.

Core Variables

Variable	Purpose	Default
`SAVANT_PORT`	Port the Svantic server listens on	`3000`
`SAVANT_BASE_URL`	Base URL for this Svantic instance	`http://localhost:3000`
`SAVANT_CENTRAL_URL`	Central hub URL (sidecar → central topology only)	—
`LLM_API_KEY`	API key for the language model provider	—
`LLM_MODEL`	Model to use for planning and execution	`gemini-2.5-flash`

Variable	Purpose	Default
`SVANTIC_CLIENT_ID`	API client ID for agent authentication	—
`SVANTIC_CLIENT_SECRET`	API client secret for agent authentication	—
`SVANTIC_INSTANCE_ID`	Stable instance ID (for containers/k8s)	hostname-port
`SAVANT_AGENTS_URL`	URL your agent uses to reach Svantic	—

Variable	Purpose	Default
`CORTEX_URL`	URL of the Cortex knowledge service	`http://localhost:3002`
`CORTEX_MONGO_URI`	MongoDB connection string for knowledge card storage	—
`EMBEDDING_MODEL`	Model used for vector embeddings	—

Variable	Purpose	Default
`DATABASE_URL`	PostgreSQL connection string for the gateway	—
`REDIS_URL`	Redis connection string for session state and caching	—

Minimal configuration — just the LLM key:

LLM_API_KEY=your-key npm run api

Same as standalone, but your application connects via SAVANT_AGENTS_URL:

SAVANT_AGENTS_URL=http://savant:3000 node my-app.js

Add SAVANT_CENTRAL_URL on each sidecar to register with the hub:

SAVANT_CENTRAL_URL=http://central-savant:3000 npm run api

The central instance does not need SAVANT_CENTRAL_URL — it is the hub.