terraform.tfvars.example

enable_provisioning    = false
deployment_name        = "vast-coding-llm"
selected_model_profile = "qwen3_coder_30b_fp8"

# Keep this at 1 for the first POC. Set to null only when ready to calculate
# capacity against target_concurrency and rent many replicas.
replica_count_override = 1
target_concurrency     = 500

# Change before provisioning. This is the key developers/tools use against vLLM.
inference_api_key = "replace-with-a-long-random-key"

market_type            = "on-demand"
max_dollars_per_hour   = 20
min_reliability        = 0.99
secure_datacenter_only = true
disk_gb                = 250
docker_image           = "vllm/vllm-openai:latest"

# Optional model profile examples:
# selected_model_profile = "glm_5_fp8"
# selected_model_profile = "glm_4_7_fp8"
# selected_model_profile = "deepseek_v3_2_exp"
# selected_model_profile = "kimi_k2_thinking"