本页尚未翻译

本页内容仍为英文原文，源自官方文档 integrations/ai/genkit。中文翻译正在分批进行中。

GenKit

genkitx-lancedb

Genkit is an open-source framework for building end-to-end AI and RAG pipelines with a clean, TypeScript-first developer experience. The genkitx-lancedb plugin lets you use LanceDB as a high-performance vector store inside your Genkit flows, so you can index, search, and retrieve data efficiently as part of your AI applications.

Installation

pnpm install genkitx-lancedb

Usage

Adding LanceDB plugin to your genkit instance.

import { lancedbIndexerRef, lancedb, lancedbRetrieverRef, WriteMode } from "genkitx-lancedb";
import { textEmbedding004, vertexAI } from "@genkit-ai/vertexai";
import { gemini } from "@genkit-ai/vertexai";
import { z, genkit } from "genkit";
import { Document } from "genkit/retriever";
import { chunk } from "llm-chunk";
import { readFile } from "fs/promises";
import path from "path";
import pdf from "pdf-parse/lib/pdf-parse";

const ai = genkit({
  plugins: [
    // vertexAI provides the textEmbedding004 embedder
    vertexAI(),

    // the local vector store requires an embedder to translate from text to vector
    lancedb([
      {
        dbUri: ".db", // optional lancedb uri, default to .db
        tableName: "table", // optional table name, default to table
        embedder: textEmbedding004,
      },
    ]),
  ],
});

You can run this app with the following command:

genkit start -- tsx --watch src/index.ts

This'll add LanceDB as a retriever and indexer to the genkit instance. You can see it in the GUI view

Testing retrieval on a sample table

To see the raw retrieval results, open the LanceDB retriever in the Genkit Developer UI, enter a query and run it:

The retriever returns 5 results from the LanceDB table. Each result looks something like this:

Creating a custom RAG flow

Now that we've seen how you can use LanceDB in a Genkit pipeline, let's refine the flow and create a RAG. A RAG flow will consist of an index and a retriever with its outputs postprocessed and fed into an LLM for final response

Creating custom indexer flows

You can also create custom indexer flows, using more options and features provided by LanceDB.

export const menuPdfIndexer = lancedbIndexerRef({
  // Using all defaults, for dbUri, tableName, and embedder, etc
});

const chunkingConfig = {
  minLength: 1000,
  maxLength: 2000,
  splitter: "sentence",
  overlap: 100,
  delimiters: "",
} as any;

async function extractTextFromPdf(filePath: string) {
  const pdfFile = path.resolve(filePath);
  const dataBuffer = await readFile(pdfFile);
  const data = await pdf(dataBuffer);
  return data.text;
}

export const indexMenu = ai.defineFlow(
  {
    name: "indexMenu",
    inputSchema: z.string().describe("PDF file path"),
    outputSchema: z.void(),
  },
  async (filePath: string) => {
    filePath = path.resolve(filePath);

    // Read the pdf.
    const pdfTxt = await ai.run("extract-text", () => extractTextFromPdf(filePath));

    // Divide the pdf text into segments.
    const chunks = await ai.run("chunk-it", async () => chunk(pdfTxt, chunkingConfig));

    // Convert chunks of text into documents to store in the index.
    const documents = chunks.map((text) => {
      return Document.fromText(text, { filePath });
    });

    // Add documents to the index.
    await ai.index({
      indexer: menuPdfIndexer,
      documents,
      options: {
        writeMode: WriteMode.Overwrite,
      } as any,
    });
  },
);

In your console, you can see the logs

Creating custom retriever flows

You can also create custom retriever flows, using more options and features provided by LanceDB.

export const menuRetriever = lancedbRetrieverRef({
  tableName: "table", // Use the same table name as the indexer.
  displayName: "Menu", // Use a custom display name.
});

export const menuQAFlow = ai.defineFlow(
  { name: "Menu", inputSchema: z.string(), outputSchema: z.string() },
  async (input: string) => {
    // retrieve relevant documents
    const docs = await ai.retrieve({
      retriever: menuRetriever,
      query: input,
      options: {
        k: 3,
      },
    });

    const extractedContent = docs.map((doc) => {
      if (doc.content && Array.isArray(doc.content) && doc.content.length > 0) {
        if (doc.content[0].media && doc.content[0].media.url) {
          return doc.content[0].media.url;
        }
      }
      return "No content found";
    });

    console.log("Extracted content:", extractedContent);

    const { text } = await ai.generate({
      model: gemini("gemini-2.0-flash"),
      prompt: `
You are acting as a helpful AI assistant that can answer 
questions about the food available on the menu at Genkit Grub Pub.

Use only the context provided to answer the question.
If you don't know, do not make up an answer.
Do not add or change items on the menu.

Context:
${extractedContent.join("\n\n")}

Question: ${input}`,
      docs,
    });

    return text;
  },
);

Now using our retrieval flow, we can ask a question about the ingested PDF