This notebook provides a quick overview for getting started with the LangSmithLoader. For detailed documentation of all LangSmithLoader features and configurations head to the API reference.

​
Overview

​
Integration details

ClassPackageLocalSerializablePY support
LangSmithLoader@langchain/community✅beta✅

​
Loader features

SourceWeb LoaderNode Envs Only
LangSmithLoader✅❌

​
Setup

To access the LangSmith document loader you’ll need to install @langchain/core, create a LangSmith account and get an API key.

​
Credentials

Sign up at langsmith.com and generate an API key. Once you’ve done this set the LANGSMITH_API_KEY environment variable: 
export LANGSMITH_API_KEY="your-api-key"

​
Installation

The LangSmithLoader integration lives in the @langchain/core package: 
import IntegrationInstallTooltip from "@mdx_components/integration_install_tooltip.mdx";
<IntegrationInstallTooltip></IntegrationInstallTooltip>

<Npm2Yarn>
  @langchain/core
</Npm2Yarn>

​
Create example dataset

For this example, we’ll create a new dataset which we’ll use in our document loader. 
import { Client as LangSmithClient } from 'langsmith';
import { faker } from "@faker-js/faker";

const lsClient = new LangSmithClient();

const datasetName = "LangSmith Few Shot Datasets Notebook";

const exampleInputs = Array.from({ length: 10 }, (_, i) => ({
  input: faker.lorem.paragraph(),
}));
const exampleOutputs = Array.from({ length: 10 }, (_, i) => ({
  output: faker.lorem.sentence(),
}));
const exampleMetadata = Array.from({ length: 10 }, (_, i) => ({
  companyCatchPhrase: faker.company.catchPhrase(),
}));

await lsClient.deleteDataset({
  datasetName,
})

const dataset = await lsClient.createDataset(datasetName);

const examples = await lsClient.createExamples({
  inputs: exampleInputs,
  outputs: exampleOutputs,
  metadata: exampleMetadata,
  datasetId: dataset.id,
});

import { LangSmithLoader } from "@langchain/core/document_loaders/langsmith"

const loader = new LangSmithLoader({
  datasetName: "LangSmith Few Shot Datasets Notebook",
  // Instead of a datasetName, you can alternatively provide a datasetId
  // datasetId: dataset.id,
  contentKey: "input",
  limit: 5,
  // formatContent: (content) => content,
  // ... other options
})

​
Load

const docs = await loader.load()
docs[0]

{
  pageContent: 'Conventus supellex aegrotatio termes. Vapulus abscido ubi vita coadunatio modi crapula comparo caecus. Acervus voluptate tergeo pariatur conor argumentum inventore vomito stella.',
  metadata: {
    id: 'f1a04800-6f7a-4232-9743-fb5d9029bf1f',
    created_at: '2024-08-20T17:01:38.984045+00:00',
    modified_at: '2024-08-20T17:01:38.984045+00:00',
    name: '#f1a0 @ LangSmith Few Shot Datasets Notebook',
    dataset_id: '9ccd66e6-e506-478c-9095-3d9e27575a89',
    source_run_id: null,
    metadata: {
      dataset_split: [Array],
      companyCatchPhrase: 'Integrated solution-oriented secured line'
    },
    inputs: {
      input: 'Conventus supellex aegrotatio termes. Vapulus abscido ubi vita coadunatio modi crapula comparo caecus. Acervus voluptate tergeo pariatur conor argumentum inventore vomito stella.'
    },
    outputs: {
      output: 'Excepturi adeptio spectaculum bis volaticus accusamus.'
    }
  }
}

console.log(docs[0].metadata)

{
  id: 'f1a04800-6f7a-4232-9743-fb5d9029bf1f',
  created_at: '2024-08-20T17:01:38.984045+00:00',
  modified_at: '2024-08-20T17:01:38.984045+00:00',
  name: '#f1a0 @ LangSmith Few Shot Datasets Notebook',
  dataset_id: '9ccd66e6-e506-478c-9095-3d9e27575a89',
  source_run_id: null,
  metadata: {
    dataset_split: [ 'base' ],
    companyCatchPhrase: 'Integrated solution-oriented secured line'
  },
  inputs: {
    input: 'Conventus supellex aegrotatio termes. Vapulus abscido ubi vita coadunatio modi crapula comparo caecus. Acervus voluptate tergeo pariatur conor argumentum inventore vomito stella.'
  },
  outputs: { output: 'Excepturi adeptio spectaculum bis volaticus accusamus.' }
}

console.log(docs[0].metadata.inputs)

{
  input: 'Conventus supellex aegrotatio termes. Vapulus abscido ubi vita coadunatio modi crapula comparo caecus. Acervus voluptate tergeo pariatur conor argumentum inventore vomito stella.'
}

console.log(docs[0].metadata.outputs)

{ output: 'Excepturi adeptio spectaculum bis volaticus accusamus.' }

console.log(Object.keys(docs[0].metadata))

[
  'id',
  'created_at',
  'modified_at',
  'name',
  'dataset_id',
  'source_run_id',
  'metadata',
  'inputs',
  'outputs'
]

​
API reference

For detailed documentation of all LangSmithLoader features and configurations head to the API reference