Configuration

All configuration is done via environment variables in your root .env file. See .env.example for the full list.

Required

Variable	Description
OPENAI_API_KEY	OpenAI (or compatible) API key
DATABASE_URL	Postgres connection string (set automatically by Docker Compose)

For browser-based runs, you also need BROWSER_USE_API_KEY.

LLM Settings

Variable	Default	Description
OPENAI_BASE_URL	https://api.openai.com/v1	Override for compatible providers
ANTHROPIC_API_KEY	--	For Claude models
TESTER_MODEL	gpt-4o-mini	Model for simulated patient
TESTER_TEMPERATURE	0.2	Temperature for tester
GRADER_MODEL	gpt-4o-mini	Model for transcript grading
GRADER_TEMPERATURE	0.1	Temperature for grader

Server

Variable	Default	Description
PORT	8000	API server port (Docker Compose maps this to 3000 externally)
NODE_ENV	development	Environment mode
WORKER_CONCURRENCY	5	Parallel scenario execution limit
DEFAULT_MAX_TURNS	11	Default number of conversation turns per scenario
MIN_MAX_TURNS	5	Minimum allowed value for max_turns
MAX_MAX_TURNS	15	Maximum allowed value for max_turns

Provider Keys

Set these based on which integrations you use:

Variable	Provider
BROWSER_USE_API_KEY	Browser Use Cloud browser testing

Per-Agent Configuration

Provider credentials like Vapi API keys, LiveKit credentials, and Pipecat API keys are configured per-agent in the agent's config object — not as server environment variables. Browser Use Cloud auth is configured with BROWSER_USE_API_KEY, while browser behavior and target URL live on the browser agent itself. See Integrations for details on each provider's config fields.

Docker Compose

These variables are used by Docker Compose but are not part of the server application config:

Variable	Default	Description
DB_PASSWORD	preclinical	Postgres password