Zum Hauptinhalt springen

Adding persistent vector storage

In the previous examples, we've been loading our data into memory each time we run the agent. This is fine for small datasets, but for larger datasets you'll want to store your embeddings in a database. LlamaIndex.TS provides a VectorStore class that can store your embeddings in a variety of databases. We're going to use Qdrant, a popular vector store, for this example.

We can get a local instance of Qdrant running very simply with Docker (make sure you install Docker first):

docker pull qdrant/qdrant
docker run -p 6333:6333 qdrant/qdrant

And in our code we initialize a VectorStore with the Qdrant URL:

// initialize qdrant vector store
const vectorStore = new QdrantVectorStore({
  url: "http://localhost:6333",
});

Now once we have loaded our documents, we can instantiate an index with the vector store:

// create a query engine from our documents
const index = await VectorStoreIndex.fromDocuments(documents, { vectorStore });

In the final iteration you can see that we have also implemented a very naive caching mechanism to avoid re-parsing the PDF each time we run the agent:

// load cache.json and parse it
let cache = {};
let cacheExists = false;
try {
  await fs.access(PARSING_CACHE, fs.constants.F_OK);
  cacheExists = true;
} catch (e) {
  console.log("No cache found");
}
if (cacheExists) {
  cache = JSON.parse(await fs.readFile(PARSING_CACHE, "utf-8"));
}

const filesToParse = ["../data/sf_budget_2023_2024.pdf"];

// load our data, reading only files we haven't seen before
let documents = [];
const reader = new LlamaParseReader({ resultType: "markdown" });
for (let file of filesToParse) {
  if (!cache[file]) {
    documents = documents.concat(await reader.loadData(file));
    cache[file] = true;
  }
}

// write the cache back to disk
await fs.writeFile(PARSING_CACHE, JSON.stringify(cache));

Since parsing a PDF can be slow, especially a large one, using the pre-parsed chunks in Qdrant can significantly speed up your agent.

Next steps​

In this guide you've learned how to

The next steps are up to you! Try creating more complex functions and query engines, and set your agent loose on the world.