L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Développement · SKILL.md

apollo-multi-env-setup

apollo-multi-env-setup

Configure Apollo.io multi-environment setup. Use when setting up development, staging, and production environments, or managing multiple Apollo configurations. Trigger with phrases like "apollo environments", "apollo staging", "apollo dev prod", "apollo multi-tenant", "apollo env config".

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: apollo-multi-env-setup description: 'Configure Apollo.io multi-environment setup.

Use when setting up development, staging, and production environments,

or managing multiple Apollo configurations.

Trigger with phrases like "apollo environments", "apollo staging",

"apollo dev prod", "apollo multi-tenant", "apollo env config".

' allowed-tools: Read, Write, Edit, Bash(kubectl:), Bash(curl:) version: 1.13.0 license: MIT author: Jeremy Longshore jeremy@intentsolutions.io tags:

  • saas
  • apollo
  • apollo-multi compatibility: Designed for Claude Code

Apollo Multi-Environment Setup

Overview

Configure Apollo.io for development, staging, and production with isolated API keys, environment-specific rate limits, feature gating, and Kubernetes-native secret management. Apollo provides sandbox tokens for testing that return dummy data without consuming credits.

Prerequisites

  • Separate Apollo API keys per environment (or sandbox tokens for dev)
  • Node.js 18+

Instructions

Step 1: Environment Configuration Schema

// src/config/apollo-config.ts
import { z } from 'zod';

const EnvironmentSchema = z.enum(['development', 'staging', 'production']);

const ApolloEnvConfig = z.object({
  environment: EnvironmentSchema,
  apiKey: z.string().min(10),
  baseUrl: z.string().url().default('https://api.apollo.io/api/v1'),
  isSandbox: z.boolean().default(false),
  rateLimit: z.object({
    maxPerMinute: z.number().min(1),
    concurrency: z.number().min(1).max(20),
  }),
  features: z.object({
    enrichment: z.boolean(),
    sequences: z.boolean(),
    deals: z.boolean(),
    bulkEnrichment: z.boolean(),
  }),
  credits: z.object({
    dailyBudget: z.number(),
    alertThreshold: z.number(),  // percentage
  }),
  logging: z.object({
    level: z.enum(['debug', 'info', 'warn', 'error']),
    redactPII: z.boolean(),
  }),
});

type ApolloEnvConfig = z.infer<typeof ApolloEnvConfig>;

Step 2: Per-Environment Configs

const configs: Record<string, ApolloEnvConfig> = {
  development: {
    environment: 'development',
    apiKey: process.env.APOLLO_SANDBOX_KEY ?? process.env.APOLLO_API_KEY_DEV!,
    baseUrl: 'https://api.apollo.io/api/v1',
    isSandbox: true,
    rateLimit: { maxPerMinute: 20, concurrency: 2 },
    features: { enrichment: true, sequences: false, deals: false, bulkEnrichment: false },
    credits: { dailyBudget: 10, alertThreshold: 80 },
    logging: { level: 'debug', redactPII: true },
  },
  staging: {
    environment: 'staging',
    apiKey: process.env.APOLLO_API_KEY_STAGING!,
    baseUrl: 'https://api.apollo.io/api/v1',
    isSandbox: false,
    rateLimit: { maxPerMinute: 50, concurrency: 5 },
    features: { enrichment: true, sequences: true, deals: true, bulkEnrichment: true },
    credits: { dailyBudget: 50, alertThreshold: 70 },
    logging: { level: 'info', redactPII: true },
  },
  production: {
    environment: 'production',
    apiKey: process.env.APOLLO_API_KEY_PROD!,
    baseUrl: 'https://api.apollo.io/api/v1',
    isSandbox: false,
    rateLimit: { maxPerMinute: 100, concurrency: 10 },
    features: { enrichment: true, sequences: true, deals: true, bulkEnrichment: true },
    credits: { dailyBudget: 500, alertThreshold: 90 },
    logging: { level: 'warn', redactPII: true },
  },
};

Step 3: Environment-Aware Client Factory

// src/config/client-factory.ts
import axios, { AxiosInstance } from 'axios';

export function createEnvClient(config: ApolloEnvConfig): AxiosInstance {
  const validated = ApolloEnvConfig.parse(config);

  const client = axios.create({
    baseURL: validated.baseUrl,
    headers: { 'Content-Type': 'application/json', 'x-api-key': validated.apiKey },
    timeout: validated.environment === 'development' ? 30_000 : 15_000,
  });

  // Feature gate: block disabled endpoints
  client.interceptors.request.use((req) => {
    if (req.url?.includes('/people/match') && !validated.features.enrichment)
      throw new Error(`Enrichment disabled in ${validated.environment}`);
    if (req.url?.includes('/emailer_campaigns') && !validated.features.sequences)
      throw new Error(`Sequences disabled in ${validated.environment}`);
    if (req.url?.includes('/opportunities') && !validated.features.deals)
      throw new Error(`Deals disabled in ${validated.environment}`);
    if (req.url?.includes('/people/bulk_match') && !validated.features.bulkEnrichment)
      throw new Error(`Bulk enrichment disabled in ${validated.environment}`);
    return req;
  });

  // Debug logging in dev
  if (validated.logging.level === 'debug') {
    client.interceptors.request.use((req) => {
      console.log(`[${validated.environment}] ${req.method?.toUpperCase()} ${req.url}`);
      return req;
    });
  }

  return client;
}

export function getClient(): AxiosInstance {
  const env = process.env.NODE_ENV ?? 'development';
  return createEnvClient(configs[env] ?? configs.development);
}

Step 4: Kubernetes Secrets

# k8s/dev/apollo-secret.yaml
apiVersion: v1
kind: Secret
metadata:
  name: apollo-credentials
  namespace: sales-dev
type: Opaque
stringData:
  APOLLO_SANDBOX_KEY: "sandbox-token-here"
  APOLLO_API_KEY_DEV: "dev-key-here"
---
# k8s/prod/apollo-secret.yaml
apiVersion: v1
kind: Secret
metadata:
  name: apollo-credentials
  namespace: sales-prod
type: Opaque
stringData:
  APOLLO_API_KEY_PROD: "master-key-here"

Step 5: Environment Verification Script

// src/scripts/verify-envs.ts
async function verifyCurrentEnvironment() {
  const env = process.env.NODE_ENV ?? 'development';
  const config = configs[env] ?? configs.development;
  try {
    const client = createEnvClient(config);
    const { data } = await client.get('/auth/health');
    const isMaster = await testMasterAccess(client);
    console.log(`${env}: ${data.is_logged_in ? 'OK' : 'FAIL'} (${isMaster ? 'master' : 'standard'} key, sandbox: ${config.isSandbox})`);
  } catch (err: any) {
    console.error(`${env}: FAIL — ${err.message}`);
  }
}

async function testMasterAccess(client: AxiosInstance): Promise<boolean> {
  try { await client.post('/contacts/search', { per_page: 1 }); return true; }
  catch { return false; }
}

Output

  • Zod-validated environment config schema with feature flags and credit budgets
  • Three environment configs (dev with sandbox, staging, production)
  • Client factory with feature gating and debug logging
  • Kubernetes secrets per namespace
  • Environment verification script testing only the active config

Examples

For a feature that adds bulk enrichment, keep the capability disabled in development and enable it first in staging with a separately injected staging credential and a low daily budget. Run the active-environment verifier from that staging deployment, confirm that disabled production-only operations are rejected in lower environments, and inspect only redacted logs. Promote the feature flag after staging telemetry is healthy. If NODE_ENV is unexpected, the active secret is missing, or a lower environment can invoke an unapproved write path, stop the release and correct the configuration rather than falling back to a production key.

Error Handling

IssueResolution
Feature disabledClient throws descriptive error identifying which env blocked it
Wrong environmentCheck NODE_ENV, client falls back to development
Missing API keyZod throws with clear validation error
Sandbox returning dummy dataExpected in development — use staging for real data testing

Resources

Next Steps

Proceed to apollo-observability for monitoring setup.

Skills enregistrées