Cloudflare Workers Vectorize Agent

We take the FAQ agent from the earlier deployment workshops and move the whole online path to Cloudflare. The Worker serves the UI and API, Workers AI embeds the query and generates the answer, and Vectorize stores the FAQ vectors.

We test whether Cloudflare can replace the usual stack for a small FAQ agent:

• FastAPI for the API server
• a separate vector database
• a paid always-on host

Cloudflare can run the full online path, with one caveat. Ingestion still runs locally as an operator task instead of becoming a public endpoint.

We start with the Cloudflare architecture decision: the deployed app stays on Workers, while ingestion runs locally as an operator command.

Then we build the Cloudflare-first path. We create credentials, create a Vectorize index, and ingest the FAQ. After that, we run the Worker locally against real Cloudflare services, deploy it, and clean it up.

In the session, we also try a Python Worker rewrite. That version is useful if you prefer Python, but Cloudflare Python Workers are still beta. The TypeScript Worker stays the stable path for the main build.

Links

Resources used here:

• End-to-End Agent Deployment - the original FastAPI FAQ agent.
• Deploying Vector Search with SQLite - the Turso and SQLite vector-search version this workshop follows.
• Cloudflare Workers - the serverless runtime.
• Cloudflare Vectorize - the managed vector index.