0 Get started
Algolia is a search engine you rent; the kit's vector capability is semantic search over content that stays in your account. They are not the same product, and for keyword search over a catalog Algolia is simply better at it. The question is which kind of search your app actually needs.
Choose Algolia
When search is a feature users judge you on.
Typo tolerance, faceting, synonyms, merchandising rules, an analytics loop that tells you what people searched for and did not find — and query latency measured in single-digit milliseconds worldwide. If your product is a catalog, that is the whole game.
Choose Pithy
When you want meaning, not matching, over content you already hold.
Workers AI embeds, Vectorize indexes, and D1 keeps the text — all inside your account. There is no document to ship to a third party, no index to keep in sync with your database, and no per-search bill.
On this capability alone
| Feature | Algolia | Pithy |
|---|---|---|
Search | ||
| Keyword search with typo tolerance Algolia: the thing it is best in the world at Pithy: not offered; this is semantic search, which is a different tool | Yes | No |
| Semantic and vector search Algolia: available, and newer than the rest of the product Pithy: the primary mode: Workers AI embeddings over Vectorize | Partial | Yes |
| Faceting, synonyms and merchandising Algolia: a deep, mature feature set Pithy: not offered | Yes | No |
| Your content stays in your account Algolia: records are indexed on their infrastructure Pithy: embeddings in your Vectorize, text in your D1 | No | Yes |
| Index kept in sync with your database Algolia: you write and operate the sync Pithy: the same migration that writes the row processes the document | Partial | Yes |
| Search analytics Algolia: a product in its own right Pithy: not offered | Yes | No |
| No per-search or per-record pricing Pithy: Cloudflare bills the embedding and the query | No | Yes |
The saffron tick marks where Pithy leads — not that the other column is absent. Every claim about Algolia was checked against their own documentation on 2026-08-25; both products change, and this page is a snapshot rather than a maintained contract.
Where Algolia wins
Distributed replicas and an engine written for exactly this. A vector query against Vectorize is fast; it is not that fast, and honest comparison says so.
Users type badly, and forgiving that well — without returning nonsense — is years of work encoded in a product.
Pinning a result, boosting a category for a campaign, running a rule for a date range. That is a commerce tool, not a search index.
Searches with no results are the most valuable data a catalog has. Algolia surfaces them; the kit leaves you to instrument it yourself.
If users type into a box and judge what comes back, buy Algolia. If what you need is to find the documents that mean the same thing as a question — for support, for retrieval, for an agent — then embedding them in your own account, without shipping your content to a third party, is the better shape and the cheaper one.