Weave Code
Ai Manticore Search Store Laravel Package
symfony/ai-manticore-search-store
ManticoreSearch Store provides a vector store integration for Symfony AI Store, enabling applications to use ManticoreSearch's KNN search for similarity retrieval. Part of the Symfony AI ecosystem.
ManticoreSearch Store provides a vector store integration for Symfony AI Store, connecting your AI applications to ManticoreSearch's fast full-text and vector search engine. It enables you to persist embeddings and run KNN (K-Nearest Neighbors) similarity searches directly through Symfony's store abstractions.
- KNN vector search for high-performance similarity queries
- Seamless integration with Symfony AI Store
- Store and retrieve embeddings alongside your documents
- Simple configuration within Symfony applications
- Backed by ManticoreSearch's scalable search engine
Ideal for building semantic search and retrieval-augmented features in Symfony projects.
Frequently asked questions about Ai Manticore Search Store
- How do I install symfony/ai-manticore-search-store in Laravel?
- Install it with Composer by requiring both symfony/ai and symfony/ai-manticore-search-store, and make sure the manticoresearch/manticoresearch PHP client is available in your project. You also need a running ManticoreSearch instance before the store can connect. In Laravel, bind the store through a service provider or facade rather than instantiating the bridge directly.
- Does this package work with Laravel or is it Symfony-only?
- It is built for Symfony AI Store, but it can be used in Laravel through Symfony's AI Store abstraction. You typically add a thin service provider or facade to expose StoreInterface to your Laravel application. At v0.14.0 the StoreFactory makes that integration lighter because construction is config or DSN-driven.
- Which Laravel and PHP versions are supported by this package?
- The package follows the Symfony AI 0.x line, and the version discussed here is v0.14.0. It requires symfony/ai and the ManticoreSearch PHP client, so your Laravel and PHP versions must satisfy their Composer constraints. Pin symfony/ai and symfony/ai-manticore-search-store to ^0.14.0 and check composer why-not before upgrading.
- How do I configure the ManticoreSearch connection in a Laravel app?
- ManticoreSearch exposes an HTTP JSON API, typically on port 9308, so the store needs the host, port, and index details for your ManticoreSearch instance. With v0.14.0, prefer the Symfony AI StoreFactory and drive construction from Laravel config or a DSN. Register the resulting store in AppServiceProvider and expose it through a facade or interface binding.
- What is VectorDocumentInterface and do I have to implement it?
- VectorDocumentInterface is the typed document contract added in the v0.14.0 line, introduced by feature #2495. It gives stores a predictable payload shape instead of relying on arrays or Eloquent models passed directly. For a greenfield integration, adopt it from the start; for an existing build, map your DTOs or models to the interface before ingestion.
- Can I use ManticoreSearch as a vector store for RAG in Laravel?
- Yes. The bridge lets Symfony AI applications use ManticoreSearch for KNN-based semantic search, which fits RAG, recommendation systems, and hybrid keyword-plus-vector search. ManticoreSearch is open source and lightweight, and its KNN support is designed for similarity search with metadata filtering. You still need to design the ManticoreSearch schema with vector fields and distance metrics upfront.
- How does KNN search work with this Symfony AI store?
- You define vector fields, indexes, and a distance metric such as L2 or cosine in ManticoreSearch before querying. The store sends nearest-neighbor requests to ManticoreSearch's HTTP API and returns matching documents through Symfony AI's StoreInterface. v0.14.0 also adds count capabilities, so you can check corpus size or ingestion results without running a full query.
- What changed in v0.14.0 and is it a breaking change?
- v0.14.0 adds StoreFactory for standardized construction, VectorDocumentInterface for typed documents, and count capabilities. The document interface is the main compatibility risk because custom document classes or array-shape assumptions must be reconciled with it. For a greenfield integration this is low risk, but for an in-flight build treat it as a compile-time and test-suite event. Pin both packages to ^0.14.0 and rerun your CRUD and KNN assertions before merging.
- How should I test a Laravel app that uses this ManticoreSearch vector store?
- Build a proof of concept that covers creating the store through the factory, indexing documents that implement VectorDocumentInterface, and running KNN queries with metadata filters. Run CRUD and KNN assertions against a real ManticoreSearch instance, because schema and client compatibility matter more than mocked HTTP responses. Pin symfony/ai and symfony/ai-manticore-search-store to ^0.14.0 so interface changes do not appear silently in CI.
- How does this compare with pgvector, Pinecone, or Elasticsearch for Laravel vector search?
- ManticoreSearch is open source and avoids the vendor lock-in of hosted services such as Pinecone or Weaviate Cloud, while providing native KNN search and filtering. If you already run PostgreSQL pgvector or Elasticsearch, moving to ManticoreSearch requires ETL to transform embeddings into ManticoreSearch's schema and document format. For v0.14.0, target VectorDocumentInterface during that migration rather than retrofitting it after ingestion.
Weaver
How can I help you explore Laravel packages today?