Add an e2e test app that uses the Vercel AI integration the way a real user does, and that sends the data to a real Sentry project.
Current state: gen-ai-libraries sends real requests through OpenRouter (Node and Cloudflare). But the SDK sends the spans to the local event proxy (tunnel: localhost:3031), so nothing gets to Sentry. cloudflare-vercelai-v7 uses MockLanguageModel and makes no real requests.
Requirements:
- Send real requests to the model provider (OpenRouter is OK). Replace the mock model in
cloudflare-vercelai-v7.
- Use a real
E2E_TEST_DSN with no tunnel, so the data goes to Sentry.
- Read the
gen_ai spans back through the Sentry API and check them, the same way the *-send-to-sentry apps do.
- Add a variant that runs against the
latest version of the library.
Add an e2e test app that uses the Vercel AI integration the way a real user does, and that sends the data to a real Sentry project.
Current state:
gen-ai-librariessends real requests through OpenRouter (Node and Cloudflare). But the SDK sends the spans to the local event proxy (tunnel: localhost:3031), so nothing gets to Sentry.cloudflare-vercelai-v7usesMockLanguageModeland makes no real requests.Requirements:
cloudflare-vercelai-v7.E2E_TEST_DSNwith no tunnel, so the data goes to Sentry.gen_aispans back through the Sentry API and check them, the same way the*-send-to-sentryapps do.latestversion of the library.