InMemoryStore

In-memory implementation of vector database storage.

This implementation stores vectors in memory using a ConcurrentHashMap for thread safety. It uses cosine similarity for vector search, which is the most common metric for embedding vectors.

This implementation is suitable for:

• Development and testing

• Small to medium-sized datasets

• Prototyping RAG systems

For production use with large datasets, consider using a dedicated vector database like Qdrant, ChromaDB, or Pinecone.

Example usage:

`InMemoryStore store = InMemoryStore.builder()
.dimensions(1024)
.build();

// Add documents
store.add(documents).block();

// Search
List<Document> results = store.search(queryEmbedding, 5, 0.8).block();
`

Exception Handling:

• IllegalArgumentException - for invalid input parameters (null documents, null embeddings, invalid limit)

• VectorStoreException - for vector-specific errors (dimension mismatch)