AI for Cloudflares
Cloudflare AI is the set of services for running machine learning from the edge. Workers AI executes hosted models through a binding, Vectorize stores embeddings for similarity search, and AI Gateway sits in front of model providers to add caching, rate limits and logging. They let you build AI features without managing GPUs.
Stage: AI and Operations · Part of the Cloudflare roadmap · Last reviewed 19 November 2026
What to learn in AI
- Call a hosted model from a Worker through the AI binding
- Generate embeddings and store them in a Vectorize index
- Build a simple retrieval flow that fetches matching documents for a prompt
- Route model calls through AI Gateway to view logs and cache responses
- Understand usage limits and costs of running inference at the edge
AI interview questions
What is a vector database used for in an AI application?
It stores embeddings, which are numeric representations of text or images, and finds the nearest ones to a query embedding. Applications use it for semantic search and retrieval-augmented generation, fetching relevant passages to include in a prompt.
What does an AI gateway add in front of model providers?
It acts as a proxy that records requests, caches repeated responses, applies rate limits, and can retry or fall back between providers. This gives visibility into cost and usage and reduces calls to the model.
Where AI fits in the Cloudflare roadmap
AI belongs to the AI and Operations stage (advanced level). Modern edge.
Other units in this stage
Back to the full Cloudflare roadmap