Built-in vector database for .NET applications.

using LMKit.Model;
using LMKit.Data;
using LMKit.Retrieval;

// Load embedding model
var embedModel = LM.LoadFromModelID("embeddinggemma-300m");

// Create in-memory DataSource
var dataSource = DataSource.CreateInMemoryDataSource("my-collection", embedModel);

// Use RagEngine to import content
var ragEngine = new RagEngine(embedModel);
ragEngine.AddDataSource(dataSource);

// Import text with automatic chunking
ragEngine.ImportText(
    "Your document content here...",
    new TextChunking() { MaxChunkSize = 500 },
    "my-collection",
    "document-section");

// Optional: Serialize to disk for later reuse
dataSource.Serialize("./cache/my-collection.bin");

// Later: Deserialize from disk
var restored = DataSource.Deserialize("./cache/my-collection.bin", embedModel);

using LMKit.Model;
using LMKit.Data;
using LMKit.Retrieval;

// Load embedding model
var embedModel = LM.LoadFromModelID("embeddinggemma-300m");

const string DATA_SOURCE_PATH = "Ebooks.dat";
const string COLLECTION_NAME = "Ebooks";

DataSource dataSource;

if (File.Exists(DATA_SOURCE_PATH))
{
    // Load existing file-based DataSource
    dataSource = DataSource.LoadFromFile(DATA_SOURCE_PATH, readOnly: false);
}
else
{
    // Create new file-based DataSource
    dataSource = DataSource.CreateFileDataSource(DATA_SOURCE_PATH, COLLECTION_NAME, embedModel);
}

// Use RagEngine to import and query
var ragEngine = new RagEngine(embedModel);
ragEngine.AddDataSource(dataSource);

// Check if section already exists
if (!dataSource.HasSection("Romeo and Juliet"))
{
    string content = File.ReadAllText("romeo_and_juliet.txt");
    ragEngine.ImportText(content,
        new TextChunking() { MaxChunkSize = 500 },
        COLLECTION_NAME, "Romeo and Juliet");
}

using LMKit.Model;
using LMKit.Data;
using LMKit.Data.Storage.Qdrant;
using LMKit.Retrieval;

// Load embedding model
var embedModel = LM.LoadFromModelID("embeddinggemma-300m");

// Connect to Qdrant (docker run -p 6333:6333 -p 6334:6334 qdrant/qdrant)
var store = new QdrantEmbeddingStore(new Uri("http://localhost:6334"));
const string COLLECTION = "Ebooks";

DataSource dataSource;

if (await store.CollectionExistsAsync(COLLECTION))
{
    // Load existing collection from Qdrant
    dataSource = DataSource.LoadFromStore(store, COLLECTION);
}
else
{
    // Create new collection in Qdrant
    dataSource = await DataSource.CreateVectorStoreDataSourceAsync(store, COLLECTION, embedModel);
}

// Use RagEngine with Qdrant-backed DataSource
var ragEngine = new RagEngine(embedModel, vectorStore: store);
ragEngine.AddDataSource(dataSource);

// Import content (automatically stored in Qdrant)
string content = await new HttpClient().GetStringAsync(
    "https://gutenberg.org/cache/epub/1513/pg1513.txt");
ragEngine.ImportText(content,
    new TextChunking() { MaxChunkSize = 500 },
    COLLECTION, "Romeo and Juliet");

using LMKit.Model;
using LMKit.Data;
using LMKit.Data.Storage;
using LMKit.Retrieval;

// Load embedding model
var embedModel = LM.LoadFromModelID("embeddinggemma-300m");

// Create FileSystemVectorStore (NEW in 2026.1.1)
// Each collection stored as separate .ds file in directory
var fsStore = new FileSystemVectorStore("./vector-collections");
const string COLLECTION = "Ebooks";

DataSource dataSource;

if (await fsStore.CollectionExistsAsync(COLLECTION))
{
    // Load existing collection (auto-cached in memory)
    dataSource = DataSource.LoadFromStore(fsStore, COLLECTION);
}
else
{
    // Create new collection
    dataSource = await DataSource.CreateVectorStoreDataSourceAsync(fsStore, COLLECTION, embedModel);
}

// FileSystemVectorStore implements IVectorStore interface
// Works with RagEngine just like Qdrant
var ragEngine = new RagEngine(embedModel);
ragEngine.AddDataSource(dataSource);

// Directory structure: ./vector-collections/Ebooks.ds
Console.WriteLine($"Store path: {fsStore.DirectoryPath}");