Skip to main content

​
Scope of the integration

If you use LlamaIndex with OpenAI or Anthropic as the LLM backend then Baserun will log API calls generating embeddings and text completions for indexing, querying, retrieval, and response synthesis. Additionally, Baserun will add logs to a trace with more information on a retrieval e.g. which nodes were selected and what score they were assigned.

​
Example

import baserun
from llama_index.core import VectorStoreIndex, SimpleDirectoryReader


@baserun.trace
def llama():
    # load documents
    documents = SimpleDirectoryReader("recipes").load_data()
    # create index
    index = VectorStoreIndex.from_documents(documents)
    # make a query
    query_engine = index.as_query_engine()
    response = query_engine.query("I have flour, sugar and butter. What am I missing if I want to bake oatmeal cookies?")
    print(response)


if __name__ == "__main__":
    baserun.init()
    llama()

import { baserun } from "baserun";
import fs from "node:fs/promises";
import {VectorStoreIndex, Document} from "llamaindex";

async function llama() {
  // load documents
  const dir = "recipes"
  const documents: Document[] = []
  for (const path of await fs.readdir(dir)) {
    const recipe = await fs.readFile(dir + "/" + path, "utf-8");
    const document = new Document({ text: recipe, id_: path });
    documents.push(document)
  }
  // create index
  const index = await VectorStoreIndex.fromDocuments(documents);
  // make a query
  const queryEngine = index.asQueryEngine();
  const response = await queryEngine.query({
    query: "I have flour, sugar and butter. What am I missing if I want to bake oatmeal cookies?",
  });
  console.log(response.toString());
}

async function main() {

  await baserun.init();

  await baserun.trace(async () => {
    await llama();
  })();
}

main();
Running this code should result in a trace looking like this:
Trace Details