Environment Variables

Configure Helix using environment variables in your .env file (Docker Compose) or Helm values (Kubernetes).

Core Configuration

Variable	Description	Default
`SERVER_URL`	Public URL where the Helix server is hosted	-
`LICENSE_KEY`	License key from deploy.helix.ml	-
`RUNNER_TOKEN`	Authentication token for Helix runners	-

Helix supports multiple AI providers. Configure one or more based on your deployment needs.

Variable	Description	Default
`INFERENCE_PROVIDER`	Primary provider: `helix`, `openai`, or `togetherai`	`helix`

Variable	Description	Default
`OPENAI_API_KEY`	OpenAI API key	-
`OPENAI_BASE_URL`	OpenAI API base URL	`https://api.openai.com/v1`
`OPENAI_MODELS`	Comma-separated list of allowed models	-

Variable	Description	Default
`TOGETHER_API_KEY`	Together AI API key	-
`TOGETHER_BASE_URL`	Together AI API base URL	`https://api.together.xyz/v1`
`TOGETHER_MODELS`	Comma-separated list of allowed models	-

Variable	Description	Default
`ANTHROPIC_API_KEY`	Anthropic API key	-
`ANTHROPIC_BASE_URL`	Anthropic API base URL	`https://api.anthropic.com/v1`
`ANTHROPIC_MODELS`	Comma-separated list of allowed models	-

Variable	Description	Default
`VLLM_API_KEY`	vLLM API key	-
`VLLM_BASE_URL`	vLLM API base URL	-
`VLLM_MODELS`	Comma-separated list of allowed models	-

Variable	Description	Default
`DYNAMIC_PROVIDERS`	Additional providers in format `provider1:api_key1:base_url1,provider2:api_key2:base_url2`	-
`ENABLE_CUSTOM_USER_PROVIDERS`	Allow users to configure their own providers	`false`
`PROVIDERS_MANAGEMENT_ENABLED`	Allow users to add their own API keys	`false`

Variable	Description	Default
`AUTH_PROVIDER`	Auth provider: `regular` (built-in) or `oidc`	`regular`
`AUTH_REGISTRATION_ENABLED`	Allow new user registration	`true`
`ADMIN_USER_IDS`	Comma-separated list of admin user IDs, or `all`	-

Variable	Description	Default
`REGULAR_AUTH_ENABLED`	Enable built-in authentication	`true`
`REGULAR_AUTH_TOKEN_VALIDITY`	Token validity duration	`168h` (7 days)
`REGULAR_AUTH_JWT_SECRET`	JWT signing secret	-

Variable	Description	Default
`TOOLS_ENABLED`	Enable agent tools	`true`
`TOOLS_MODEL`	Model for tool execution	`llama3:instruct`
`TOOLS_TLS_SKIP_VERIFY`	Skip TLS verification for tool requests	`false`

Variable	Description	Default
`HELIX_MODEL_TTL`	How long to keep models warm	`10s`
`HELIX_SLOT_TTL`	Time before slots are considered dead	`600s`
`HELIX_RUNNER_TTL`	Time before runners are considered dead	`30s`
`HELIX_SCHEDULING_STRATEGY`	Scheduling strategy: `max_spread` or `bin_pack`	`max_spread`
`HELIX_QUEUE_SIZE`	Workload queue buffer size	`100`

Variable	Description	Default
`APP_URL`	Public app URL for notification links	`https://app.helix.ml`
`EMAIL_SMTP_HOST`	SMTP server hostname	-
`EMAIL_SMTP_PORT`	SMTP server port	-
`EMAIL_SMTP_USERNAME`	SMTP username	-
`EMAIL_SMTP_PASSWORD`	SMTP password	-
`EMAIL_MAILGUN_DOMAIN`	Mailgun domain	-
`EMAIL_MAILGUN_API_KEY`	Mailgun API key	-

Configure runners using these environment variables (see runner_config.go):

Variable	Description	Default
`RUNTIME_OLLAMA_ENABLED`	Enable Ollama LLM inference	`true`
`RUNTIME_OLLAMA_WARMUP_MODELS`	Models to pre-download (comma-separated)	-
`RUNTIME_AXOLOTL_ENABLED`	Enable fine-tuning runtime	`true`
`RUNTIME_AXOLOTL_WARMUP_MODELS`	Fine-tuning models to pre-warm	-
`HF_TOKEN`	HuggingFace token for model access	-

Example warmup models:

RUNTIME_OLLAMA_WARMUP_MODELS=llama3:instruct,qwen3:8b,mixtral:instruct

Last updated on December 19, 2025