# Copy this file to .env and add your OpenAI API key. # --- Required ---------------------------------------------------------------- # One key runs everything: the agent, web search, image generation, voice. OPENAI_API_KEY= # docker compose sets this for you. For `pnpm dev`, point it at any Postgres 14+. DATABASE_URL=postgres://amuse:amuse@localhost:5432/amuse # --- Optional ---------------------------------------------------------------- # Models (defaults shown). # AMUSE_AGENT_MODEL=gpt-6.2-sol # AMUSE_FAST_MODEL=gpt-6-luna # AMUSE_REASONING_EFFORT=low # low | medium | high # AMUSE_IMAGE_MODEL=gpt-image-2 # AMUSE_TRANSCRIBE_MODEL=gpt-4o-mini-transcribe # AMUSE_SPEECH_MODEL=gpt-4o-mini-tts # An OpenAI-compatible endpoint or proxy. # OPENAI_BASE_URL= # Daily spend cap in usage units (about 1 unit per 1k input tokens). Default 2000. # AMUSE_DAILY_LIMIT=2000 # Server port (default 3000). In dev, open the web app on http://localhost:5173. # PORT=3000 # Address to listen on (default 127.0.0.1). Amuse has no login: anything other than a # loopback address lets other machines control it. # HOST=127.0.0.1 # Chromium for the agent's live browser. `pnpm setup:browser` installs one; # set this to use a Chrome/Chromium you already have. # CHROMIUM_PATH= # AMUSE_BROWSER=0 # turn the live browser off # Chromium uses its own sandbox when it can, and falls back without it (as root or in Docker). # AMUSE_BROWSER_REQUIRE_SANDBOX=1 # fail instead of falling back # AMUSE_BROWSER_NO_SANDBOX=1 # skip the sandbox attempt # Where Amuse keeps the encryption key for saved credentials (default ./.data). # Back this folder up together with your database. # AMUSE_DATA_DIR=