FMXExpress has released LocalVector, a native Delphi library for building semantic search, vector databases, and Retrieval-Augmented Generation (RAG) applications entirely on local infrastructure. It actually generates the embedding vectors locally.
LocalVector combines ONNX Runtime, SQLite, and sqlite-vec to provide embedding generation, vector storage, and similarity search in a lightweight Delphi-friendly package with no external vector database required.
The framework includes support for:
- Local vector storage and search
- Semantic similarity retrieval
- RAG application development
- SQLite and sqlite-vec integration
- ONNX Runtime inference
- Local-first AI architectures
LocalVector ships with three built-in embedding models:
- all-MiniLM-L6-v2
- BAAI/bge-small-en-v1.5
- mixedbread-ai/mxbai-embed-large-v1
Whether you’re building AI assistants, agent memory systems, document search, or offline RAG workflows, LocalVector provides the core retrieval layer needed to bring semantic search to Delphi applications.
Explore LocalVector’s built-in embeddings, vector search engine, and RAG capabilities for Delphi.