Create a Cloudflare AI Search (formerly AutoRAG) instance over an R2 bucket and query it from a Worker binding or the REST API

Documented steps

1. Confirm the current naming before writing any code: the product is now 'Cloudflare AI Search' and the docs root is https://developers.cloudflare.com/ai-search/ . It was previously called AutoRAG.
2. Create and populate an R2 bucket with the source documents you want indexed.
3. Create the instance with Wrangler: wrangler ai-search create my-instance --type r2 --source my-bucket (https://developers.cloudflare.com/ai-search/get-started/wrangler/), or via Dashboard > AI Search > Create with R2 as data source.
4. Wrangler auto-generates the service API token the instance needs; if creating via REST or dashboard, generate it as described at https://developers.cloudflare.com/ai-search/configuration/data-source/r2/ .
5. AI Search indexes the bucket on a schedule. The default sync interval is 6 hours and can be set to 1, 2, 4, 6, 12 or 24 hours via the sync_interval field.
6. Bind it in wrangler.toml as a namespace binding: [[ai_search_namespaces]] binding = 'AI_SEARCH' / namespace = 'my-namespace' — or bind a single instance directly: [[ai_search]] binding = 'MY_SEARCH' / instance_name = 'my-instance'.
7. Query for chunks from a Worker via the namespace binding: const instance = env.AI_SEARCH.get('my-instance'); const results = await instance.search({ messages: [{ role: 'user', content: query }] });
8. With a direct instance binding the call is one level shallower: const results = await env.MY_SEARCH.search({ messages: [{ role: 'user', content: query }] });
9. For a generated answer rather than raw chunks, call chatCompletions({ messages: [...] }) on the same binding.
10. REST search: POST https://api.cloudflare.com/client/v4/accounts/{account_id}/ai-search/instances/{instance}/search with 'Authorization: Bearer {token}' and body {"messages":[{"role":"user","content":"What is Cloudflare?"}]}. The token needs AI Search Edit and Run permissions.
11. REST generated answer: POST .../ai-search/instances/{instance}/chat/completions with the same messages body; add "stream": true for streaming.
12. Check indexing progress with: wrangler ai-search stats my-instance