Imported from jeremylongshore/tons-of-skills-marketplace (skills/.curated/castai-sdk-patterns/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill castai-sdk-patterns. Copyright stays with the author (MIT).

CAST AI SDK Patterns

Overview

CAST AI uses a REST API with X-API-Key header authentication. There is no official SDK -- build typed wrappers around fetch or requests. These patterns cover singleton clients, typed responses, retry with backoff, and multi-cluster management.

Prerequisites

• Completed castai-install-auth setup
• TypeScript 5+ or Python 3.10+
• Familiarity with async/await patterns

Instructions

Step 1: TypeScript API Client

// src/castai/client.ts
interface CastAIConfig {
  apiKey: string;
  baseUrl?: string;
  timeoutMs?: number;
}

interface CastAICluster {
  id: string;
  name: string;
  status: string;
  providerType: "eks" | "gke" | "aks";
  agentStatus: string;
  createdAt: string;
}

interface CastAISavings {
  monthlySavings: number;
  savingsPercentage: number;
  currentMonthlyCost: number;
  optimizedMonthlyCost: number;
}

interface CastAINode {
  name: string;
  instanceType: string;
  lifecycle: "on-demand" | "spot";
  allocatableCpu: string;
  allocatableMemory: string;
  zone: string;
}

class CastAIClient {
  private apiKey: string;
  private baseUrl: string;
  private timeoutMs: number;

  constructor(config: CastAIConfig) {
    this.apiKey = config.apiKey;
    this.baseUrl = config.baseUrl ?? "https://api.cast.ai";
    this.timeoutMs = config.timeoutMs ?? 30000;
  }

  private async request<T>(path: string, options?: RequestInit): Promise<T> {
    const controller = new AbortController();
    const timeout = setTimeout(() => controller.abort(), this.timeoutMs);

    try {
      const response = await fetch(`${this.baseUrl}${path}`, {
        ...options,
        headers: {
          "X-API-Key": this.apiKey,
          "Content-Type": "application/json",
          ...options?.headers,
        },
        signal: controller.signal,
      });

      if (!response.ok) {
        const body = await response.text();
        throw new CastAIError(response.status, body, path);
      }

      return response.json();
    } finally {
      clearTimeout(timeout);
    }
  }

  async listClusters(): Promise<CastAICluster[]> {
    const data = await this.request<{ items: CastAICluster[] }>(
      "/v1/kubernetes/external-clusters"
    );
    return data.items;
  }

  async getSavings(clusterId: string): Promise<CastAISavings> {
    return this.request(`/v1/kubernetes/clusters/${clusterId}/savings`);
  }

  async listNodes(clusterId: string): Promise<CastAINode[]> {
    const data = await this.request<{ items: CastAINode[] }>(
      `/v1/kubernetes/external-clusters/${clusterId}/nodes`
    );
    return data.items;
  }

  async updatePolicies(clusterId: string, policies: Record<string, unknown>): Promise<void> {
    await this.request(`/v1/kubernetes/clusters/${clusterId}/policies`, {
      method: "PUT",
      body: JSON.stringify(policies),
    });
  }
}

class CastAIError extends Error {
  constructor(
    public readonly status: number,
    public readonly body: string,
    public readonly path: string
  ) {
    super(`CAST AI ${status} on ${path}: ${body}`);
    this.name = "CastAIError";
  }

  get retryable(): boolean {
    return this.status === 429 || this.status >= 500;
  }
}

Step 2: Singleton with Retry

// src/castai/index.ts
let instance: CastAIClient | null = null;

export function getCastAIClient(): CastAIClient {
  if (!instance) {
    if (!process.env.CASTAI_API_KEY) {
      throw new Error("CASTAI_API_KEY environment variable required");
    }
    instance = new CastAIClient({ apiKey: process.env.CASTAI_API_KEY });
  }
  return instance;
}

export async function withRetry<T>(
  fn: () => Promise<T>,
  maxRetries = 3
): Promise<T> {
  for (let attempt = 0; attempt <= maxRetries; attempt++) {
    try {
      return await fn();
    } catch (err) {
      if (attempt === maxRetries) throw err;
      if (err instanceof CastAIError && !err.retryable) throw err;
      const delay = 1000 * Math.pow(2, attempt) + Math.random() * 500;
      await new Promise((r) => setTimeout(r, delay));
    }
  }
  throw new Error("Unreachable");
}

Step 3: Python Client

# castai_client.py
import os
import time
import requests
from dataclasses import dataclass
from typing import Optional

@dataclass
class CastAIConfig:
    api_key: str
    base_url: str = "https://api.cast.ai"
    timeout: int = 30

class CastAIClient:
    def __init__(self, config: Optional[CastAIConfig] = None):
        self.config = config or CastAIConfig(
            api_key=os.environ["CASTAI_API_KEY"]
        )
        self.session = requests.Session()
        self.session.headers.update({
            "X-API-Key": self.config.api_key,
            "Content-Type": "application/json",
        })

    def _get(self, path: str) -> dict:
        resp = self.session.get(
            f"{self.config.base_url}{path}",
            timeout=self.config.timeout,
        )
        resp.raise_for_status()
        return resp.json()

    def list_clusters(self) -> list[dict]:
        return self._get("/v1/kubernetes/external-clusters")["items"]

    def get_savings(self, cluster_id: str) -> dict:
        return self._get(f"/v1/kubernetes/clusters/{cluster_id}/savings")

    def list_nodes(self, cluster_id: str) -> list[dict]:
        return self._get(
            f"/v1/kubernetes/external-clusters/{cluster_id}/nodes"
        )["items"]

    def get_policies(self, cluster_id: str) -> dict:
        return self._get(f"/v1/kubernetes/clusters/{cluster_id}/policies")

Error Handling

Output

Provide a small client boundary that returns typed, redacted domain data or a classified failure with request correlation information. The boundary must not log API keys, full authorization headers, or unfiltered provider responses; callers need an explicit retry policy and a safe error category rather than a raw exception string.

Examples

Call list_clusters() with a read-only development key and expose only the cluster ID and agent status to the caller. When the provider returns 429, capture the response category and retry according to the bounded backoff policy; on 401, stop retries and route the request to the secret-owner rotation process.

Resources

• CAST AI OpenAPI Spec
• CAST AI Terraform Provider Source

Next Steps

Apply these patterns in castai-core-workflow-a to manage cluster optimization.