variables.tf

# =============================================================================
# OpenStack Configuration
# =============================================================================

variable "openstack_cloud" {
  type        = string
  default     = "openstack"
  description = "Name of the cloud in your clouds.yaml (typically 'openstack')"
}

variable "prefix" {
  type        = string
  default     = ""
  description = "Prefix all resources to prevent name collisions"
}

variable "vm_total" {
  type        = number
  default     = 1
  description = "Total of vms to be created"
}

variable "vm_flavor_name" {
  type        = string
  default     = "eg1.a100x1.V12-84"
  description = "Flavor type of the VM (default: 1x A100 80GB)"
}

variable "vm_additional_storage" {
  type        = number
  default     = 200
  description = "Additional storage volume in gigabytes (for model cache)"
}

variable "subnet_cidr" {
  type        = string
  default     = "192.168.42.0/24"
  description = "Subnet cidr to be used"
}

variable "vm_cidr_whitelist" {
  type        = string
  default     = "0.0.0.0/0"
  description = "Whitelist source ip to access the VM (recommended: restrict to your office/VPN IP)"
}

# =============================================================================
# vLLM Configuration
# =============================================================================

variable "vllm_model" {
  type        = string
  default     = "Qwen/Qwen3-Coder-30B-A3B-Instruct"
  description = "HuggingFace model ID to serve (default: Qwen3-Coder optimized for agentic coding)"
}

variable "huggingface_token" {
  type        = string
  default     = ""
  sensitive   = true
  description = "HuggingFace API token for downloading gated models (get one at https://huggingface.co/settings/tokens)"
}

variable "vllm_api_key" {
  type        = string
  default     = ""
  sensitive   = true
  description = "API key for vLLM authentication (auto-generated if empty)"
}

variable "vllm_max_model_len" {
  type        = number
  default     = 32768
  description = "Maximum context length for the model (reduce if OOM, increase for longer context)"
}

variable "vllm_gpu_memory_utilization" {
  type        = number
  default     = 0.9
  description = "Fraction of GPU memory to use (0.0-1.0)"
}