bernardobrugg/neo-skills-set/lancedb

npm install @lancedb/lancedb
# Optional: for automatic embedding generation
npm install @lancedb/lancedb openai

// db.ts — Create a LanceDB table and search
import * as lancedb from "@lancedb/lancedb";
// Connect to local database (creates directory if needed)
const db = await lancedb.connect("./my-vector-db");
// Create a table with data
const data = [
  { id: 1, text: "The cat sat on the mat", vector: [0.1, 0.2, 0.3, ...] },
  { id: 2, text: "Dogs are loyal companions", vector: [0.4, 0.5, 0.6, ...] },
  { id: 3, text: "Fish swim in the ocean", vector: [0.7, 0.8, 0.9, ...] },
];
const table = await db.createTable("documents", data);
// Vector search — find similar items
const results = await table
  .vectorSearch([0.1, 0.2, 0.3, ...])  // Query vector
  .limit(5)
  .toArray();
// results: [{ id: 1, text: "The cat sat on the mat", _distance: 0.001 }, ...]

// auto-embed.ts — LanceDB generates embeddings automatically
import * as lancedb from "@lancedb/lancedb";
import { getRegistry } from "@lancedb/lancedb/embeddings";
const openai = getRegistry().get("openai")!.create({
  model: "text-embedding-3-small",
});
const db = await lancedb.connect("./my-db");
// Define schema with embedding function
const schema = lancedb
  .schema([
    lancedb.field("id", new lancedb.Int32()),
    lancedb.field("text", new lancedb.Utf8(), openai.sourceField()),
    lancedb.field("vector", openai.vectorField()),  // Auto-generated
  ]);
const table = await db.createTable("docs", [
  { id: 1, text: "How to set up authentication" },
  { id: 2, text: "Database migration guide" },
  { id: 3, text: "Deploying to production" },
], { schema });
// Search with text — embedding generated automatically
const results = await table
  .search("how do I deploy my app?")
  .limit(3)
  .toArray();

// hybrid.ts — Combine keyword and semantic search
const table = await db.openTable("documents");
// Create full-text search index
await table.createIndex("text", { config: lancedb.Index.fts() });
// Hybrid search: combines vector similarity + keyword matching
const results = await table
  .search("deploy production", { queryType: "hybrid" })
  .limit(10)
  .toArray();

// filter.ts — Vector search with metadata filters
const results = await table
  .vectorSearch(queryVector)
  .where("category = 'docs' AND created_at > '2026-01-01'")
  .limit(10)
  .toArray();