env.sample

AI_GATEWAY_API_KEY=

# Optional model overrides. Values must be Vercel AI Gateway model ids.
EVEABLE_ROOT_MODEL=openai/gpt-5.4-mini
INTENT_AGENT_MODEL=openai/gpt-5.4-mini
ORCHESTRATOR_AGENT_MODEL=openai/gpt-5.4-mini
DESIGN_RESEARCH_AGENT_MODEL=openai/gpt-5.4-mini
CODE_WRITER_AGENT_MODEL=openai/gpt-5.4
AUTOFIX_AGENT_MODEL=openai/gpt-5.4-mini
SECURITY_REVIEW_AGENT_MODEL=openai/gpt-5.4-mini
CONVERSATION_AGENT_MODEL=openai/gpt-5.4-mini

# Optional image search for generated applications.
UNSPLASH_ACCESS_KEY=
UNSPLASH_API_BASE_URL=https://api.unsplash.com

# Optional Refero MCP design inspiration for the design_research subagent.
# REFERO_API_KEY is preferred; REFERO_MCP_TOKEN is accepted as a fallback.
REFERO_MCP_URL=https://api.refero.design/mcp
REFERO_API_KEY=
REFERO_MCP_TOKEN=

# Optional server-only placeholders for generated apps that need InsForge.
INSFORGE_API_BASE_URL=
INSFORGE_API_KEY=

# Required when Eveable should deploy generated apps to Vercel.
VERCEL_TOKEN=

# Optional deployment controls.
VERCEL_PROJECT_NAME=
VERCEL_SCOPE=

# Comma-separated server-only env vars to pass through Vercel deployment flags.
# NEXT_PUBLIC_* and token vars are rejected by the deploy tool.
EVEABLE_DEPLOY_ENV_ALLOWLIST=