@@ -4,10 +4,24 @@ Provides the RunPod (GPU) cloud provider adapter for sh1pt scale and deploy work
44
55## What it does
66
7- - Connects cloud provider credentials and project settings.
8- - Supports infrastructure planning, deployment, or status workflows where implemented.
9- - Includes a connection flow for account or credential setup.
10- - Includes setup guidance for required credentials or provider configuration.
7+ - Connects to RunPod through the GraphQL API with ` RUNPOD_API_KEY ` .
8+ - Quotes GPU pods from either explicit ` hourlyPrice ` config or RunPod ` gpuTypes ` pricing.
9+ - Uses the highest available price when ` cloudType ` is ` ALL ` so ` maxHourlyPrice ` is not checked against a lower community-only estimate.
10+ - Uses on-demand GPU prices for guardrails; spot bids are not sent by this adapter.
11+ - Fails clearly when the requested GPU type is not returned instead of silently selecting another GPU.
12+ - Provisions on-demand GPU pods with ` podFindAndDeployOnDemand ` .
13+ - Lists account pods, checks a pod by ID, and terminates pods with ` podTerminate ` .
14+ - Requires ` imageName ` for real provisioning and supports ` maxHourlyPrice ` guardrails before any pod is created.
15+ - Does not call RunPod during dry-run provisioning; set ` hourlyPrice ` to simulate a non-zero cost in dry-run output.
16+
17+ Common config fields:
18+
19+ - ` gpuTypeId ` : RunPod GPU type ID, such as ` NVIDIA RTX A6000 ` .
20+ - ` imageName ` : container image for real pod creation, such as ` runpod/pytorch ` .
21+ - ` cloudType ` : ` ALL ` , ` COMMUNITY ` , or ` SECURE ` .
22+ - ` hourlyPrice ` : optional explicit offline quote value per GPU.
23+ - ` volumeInGb ` : optional network volume size; omitted unless explicitly requested through config or the spec storage field.
24+ - ` ports ` , ` dockerArgs ` , ` containerDiskInGb ` , ` minVcpuCount ` , ` minMemoryInGb ` , ` volumeMountPath ` , ` networkVolumeId ` , and ` env ` .
1125
1226## Package
1327
0 commit comments