Cloudflare AI Search can now summarise files as it indexes them
An opt-in flag has AI Search run each document through an LLM during indexing, generating a summary, topics and keywords to improve retrieval.
Cloudflare’s AI Search, its managed retrieval-augmented-generation platform, can now generate a summary, topics and keywords for a file while indexing it. For any text document that produces at least three chunks, AI Search runs the content through an LLM: the summary and keywords go into the item’s metadata, and the topics get added to the context header of that file’s chunks to help retrieval.
The feature is off by default. You turn it on from the dashboard, or by setting indexing_options.file_summary to true through the API. Switching it on triggers a full reindex, and Cloudflare warns that indexing_options gets replaced wholesale on every update, so you need to pass every current option each time or the unspecified ones revert to defaults.
Why it matters: if you’re building a RAG pipeline on AI Search, this removes the need to run your own summarisation step before indexing, and the added topic context should help retrieval on long or unevenly-written documents.
The caveat: the LLM calls are billed separately as Workers AI usage, on top of your normal AI Search allowance, so turning this on for a large existing corpus is not free.