Meilisearch

Last updated: Nov 4, 2026

---

Rationale

Fluid Attacks uses Meilisearch Cloud as one search engine for keyword and semantic search in the documentation of Fluid Attacks. Thus, Fluid Attacks does not have to maintain a different provider for each search type.

With Meilisearch, Fluid Attacks gets these results:

• It supports hybrid search that automatically balances keyword and vector results for each query.
• It makes embeddings automatically through auto-embedding. Thus, Fluid Attacks does not have to manage embedding infrastructure.
• It supports index replacement: it clears and adds all the documents again on each reindex. It also supports atomic swap if it becomes necessary in the future.
• It highlights and crops search results natively, with control of the snippets for each field.
• It supports filtering and faceting. Fluid Attacks uses this function for access control (public or private documents).
• It gives a full CRUD API for document management. Thus, Fluid Attacks can audit and list the indexed content.
• It has one RESTful API that integrates with the docs site and with MCP-based consumers, for example, the MCP server of Fluid Attacks.
• It costs about $30 each month on the Build plan, with AI and semantic search included.
• It has the MIT license.

Alternatives

Fluid Attacks used these alternatives before, and Meilisearch replaced them.

Algolia

Algolia is a search-as-a-service provider that is a standard in the industry.

• Its standard plans do not include semantic search. NeuralSearch is only in the Elevate tier, for more than $50,000 each year. Without it, a second provider is necessary for semantic search.
• It does not make embeddings, because semantic search is available only on the Elevate tier.
• It supports atomic index replacement with replaceAllObjects in one call.
• It gives native highlighting for each field.
• It gives powerful faceting and configurable searchable attributes.
• It gives a full CRUD API for document management.
• It has a RESTful API with very good SDKs, but only for keyword search. Thus, MCP-based semantic consumers must have a different provider.
• It costs about $50 each month for keyword-only search at the usage level of Fluid Attacks.
• It has a proprietary license.

AWS Kendra

AWS Kendra is an enterprise search service of AWS.

• It supports keyword search with QueryCommand and semantic search with RetrieveCommand, but it does not combine them in one hybrid query.
• It does not support configurable embedding models. It uses its own built-in retrieval model.
• It does not support atomic index replacement. To replace the index content, it is necessary to rotate the data sources as a workaround.
• It does not support highlighting on semantic search results.
• It supports only basic filtering with AttributeFilter on document attributes, and it does not give faceting.
• It does not give a document listing API. Thus, it is not possible to list or audit the content of the index.
• It has an API based on the AWS SDK, but keyword and semantic search are different operations. Thus, each type of search uses a different API call.
• It costs about $300 each month at the usage level of Fluid Attacks.
• It has a proprietary license.

Typesense Cloud

Typesense is an open-source search engine with a cloud service.

• It supports hybrid keyword and vector search.
• It makes embeddings automatically with auto-embedding and optional GPU acceleration.
• It supports atomic index replacement with collection aliases, but this is a manual process with three steps.
• It gives native highlighting for each field.
• It supports dynamic filtering and faceting at query time.
• It gives a full CRUD API for document management.
• It has a RESTful API with Algolia-compatible clients that can serve keyword and semantic consumers.
• It costs about $30 each month, but the user must decide the cluster capacity manually.
• It has the GPL-3.0 license.

Usage

Fluid Attacks uses Meilisearch Cloud as the centralized search engine for all its documentation search. It is the engine of the search feature on docs.fluidattacks.com and of the semantic retrieval for the MCP server.