TrainLoop SDK

This is the SDK for TrainLoop. It allows you to send data to TrainLoop to be used for finetuning.

Installation

Python

pip install trainloop-sdk

Go

go get -u github.com/TrainLoop/[email protected]

TypeScript

npm install @trainloop/sdk

Getting an API key

You can get an API key from the TrainLoop dashboard.

Usage

Python

import os
from trainloop import Client, SampleFeedbackType
from dotenv import load_dotenv

# Load environment variables from .env file
load_dotenv()

def main():
    # Initialize the TrainLoop client with your api key
    client = Client(api_key=os.getenv("TRAINLOOP_API_KEY"))

    # Some example message thread - this should be generated by your system
    messages = [
        {"role": "system", "content": "System message here"},
        {"role": "user", "content": "Hello from the user!"},
    ]

    # Send data to trainloop and let us know if this was a success case or a failure case
    success = client.send_data(
        messages, sample_feedback=SampleFeedbackType.GOOD, dataset_id="test-dataset"
    )
    if success:
        print("Data sent successfully!")
    else:
        print("Data was not sent. Check logs for details.")


if __name__ == "__main__":
    main()

Go

package main

// 1. Import the SDK
import (
	trainloop "github.com/TrainLoop/sdk/go"
)

// 2. Get the api key from the environment
apiKey := os.Getenv("TRAINLOOP_API_KEY")

if apiKey == "" {
    log.Fatal("TRAINLOOP_API_KEY environment variable is not set")
}

// 3. Initialize the SDK
client := trainloop.NewClient(apiKey)

// 4. Do some inference and generate some message thread
messages := []trainloop.Message{
    {Role: "user", Content: "Hello, from the user"},
    {Role: "assistant", Content: "Hello, from the assistant"},
}

// 3. Send data to TrainLoop (good or bad)
success, err := client.SendData(messages, trainloop.GoodFeedback, "example-dataset-id")
if err != nil {
    log.Printf("Failed to send data to TrainLoop: %v", err)
    return
}

if success {
    log.Printf("Data sent to TrainLoop successfully")
}

TypeScript

import { Client, SampleFeedbackType } from '@trainloop/sdk';
import * as dotenv from 'dotenv';

async function main() {
  dotenv.config();
  
  // Check if API key was loaded
  const apiKey = process.env.TRAINLOOP_API_KEY;
  if (!apiKey) {
    console.error('API key not found in environment variables');
    return;
  }
  
  // Initialize the TrainLoop client with your api key
  const client = new Client(apiKey);

  // Example messages
  const messages = [
    { role: 'system', content: 'System message here' },
    { role: 'user', content: 'Hello from the user!' },
  ];

  // Send data
  const success = await client.sendData(
    messages,
    SampleFeedbackType.GOOD,
    'test-dataset'
  );

  if (success) {
    console.log('Data sent successfully!');
  } else {
    console.log('Data was not sent. Check logs for details.');
  }
}

main().catch(error => {
  console.error('Unexpected error:', error);
});

Environment Setup

For all SDKs, it's recommended to use environment variables to manage your API key. Create a .env file in your project:

TRAINLOOP_API_KEY=your_api_key_here

Each example shows how you can load from the environment variables.

• Python: Use python-dotenv with load_dotenv()
• TypeScript/JavaScript: Use dotenv with dotenv.config()
• Go: Environment variables are loaded automatically by the operating system or your deployment platform