@logtura/driver-cloudflare-ai-gateway

Logtura provider driver for Cloudflare AI Gateway. Polls GET /accounts/<id>/ai-gateway/gateways/<gw>/logs over Vector's http_client source. One source per selected gateway, 30s poll cadence.

npm install @logtura/driver-cloudflare-ai-gateway @logtura/core

Credentials

A Cloudflare API token with AI Gateway: Read scoped to the account that owns the gateway. Create one at dash.cloudflare.com/profile/api-tokens.

Usage

import { generateBundle } from "@logtura/core";
import { cloudflareAiGatewayDriver } from "@logtura/driver-cloudflare-ai-gateway";

const accounts = await cloudflareAiGatewayDriver.verifyCredentials({
  apiToken: process.env.CLOUDFLARE_API_TOKEN!,
});
const gateways = await cloudflareAiGatewayDriver.discoverSources({
  credentials: { apiToken: process.env.CLOUDFLARE_API_TOKEN! },
  accountId: accounts[0].id,
});

const bundle = generateBundle({
  providers: [cloudflareAiGatewayDriver],
  destinations: [/* ... */],
  connections: [{
    connection: {
      id: "con_a", provider: "cloudflare-ai-gateway",
      displayName: "prod", externalAccountId: accounts[0].id,
    },
    selectedSources: gateways,
    credentials: { apiToken: process.env.CLOUDFLARE_API_TOKEN! },
  }],
  monitors: [/* ... */],
});

What it emits

Per selected gateway, one Vector http_client source polling Cloudflare's AI Gateway logs endpoint. The driver's normalize maps the response (provider, model, status_code, success) into the uniform { .script, .message, .level, .error } shape. .level becomes error when !success || status >= 500. .script is the gateway's provider slug (openai, anthropic, and similar).

Runtime requirements

None beyond Vector itself. The http_client source is built in.

License

Apache 2.0.