Home

Awesome

Ballerina OpenAI Finetunes connector

Build Trivy GraalVM Check GitHub Last Commit GitHub Issues

Overview

OpenAI, an AI research organization focused on creating friendly AI for humanity, offers the OpenAI API to access its powerful AI models for tasks like natural language processing and image generation.

The ballarinax/openai.finetunes package offers APIs to connect and interact with the fine-tuning related endpoints of OpenAI REST API v1 allowing users to customize OpenAI's AI models to meet specific needs.

Setup guide

To use the OpenAI Connector, you must have access to the OpenAI API through a OpenAI Platform account and a project under it. If you do not have a OpenAI Platform account, you can sign up for one here.

Create a OpenAI API Key

  1. Open the OpenAI Platform Dashboard.

  2. Navigate to Dashboard -> API keys <img src=https://raw.githubusercontent.com/ballerina-platform/module-ballerinax-openai.finetunes/main/docs/setup/resources/navigate-api-key-dashboard.png alt="OpenAI Platform" style="width: 70%;">

  3. Click on the "Create new secret key" button <img src=https://raw.githubusercontent.com/ballerina-platform/module-ballerinax-openai.finetunes/main/docs/setup/resources/api-key-dashboard.png alt="OpenAI Platform" style="width: 70%;">

  4. Fill the details and click on Create secret key <img src=https://raw.githubusercontent.com/ballerina-platform/module-ballerinax-openai.finetunes/main/docs/setup/resources/create-new-secret-key.png alt="OpenAI Platform" style="width: 70%;">

  5. Store the API key securely to use in your application <img src=https://raw.githubusercontent.com/ballerina-platform/module-ballerinax-openai.finetunes/main/docs/setup/resources/saved-key.png alt="OpenAI Platform" style="width: 70%;">

Quickstart

To use the OpenAI Finetunes connector in your Ballerina application, update the .bal file as follows:

Step 1: Import the module

Import the openai.finetunes module.

import ballerinax/openai.finetunes;
import ballerina/io;

Step 2: Instantiate a new connector

Create a finetunes:ConnectionConfig with the obtained API Key and initialize the connector.

configurable string token = ?;

final finetunes:Client openAIFinetunes = check new({
    auth: {
        token
    }
});

Step 3: Invoke the connector operation

Now, utilize the available connector operations.

Note: First, create a sample.jsonl file in the same directory. This file should contain the training data formatted according to the guidelines provided here.

Fine tuning the gpt-3.5-turbo model

public function main() returns error? {

    finetunes:CreateFileRequest req = {
        file: {fileContent: check io:fileReadBytes("sample.jsonl"), fileName: "sample.jsonl"},
        purpose: "fine-tune"
    };

    finetunes:OpenAIFile fileRes = check openAIFinetunes->/files.post(req);

    string fileId = fileRes.id;

    finetunes:CreateFineTuningJobRequest fineTuneRequest = {
        model: "gpt-3.5-turbo",
        training_file: fileId
    };

    finetunes:FineTuningJob fineTuneResponse = 
        check openAIFinetunes->/fine_tuning/jobs.post(fineTuneRequest);
}

Step 4: Run the Ballerina application

bal run

Examples

The OpenAI Finetunes connector provides practical examples illustrating usage in various scenarios. Explore these examples, covering the following use cases:

  1. Sarcastic bot - Fine-tune the GPT-3.5-turbo model to generate sarcastic responses

  2. Sports headline analyzer - Fine-tune the GPT-4o-mini model to extract structured information (player, team, sport, and gender) from sports headlines.

Build from the source

Setting up the prerequisites

  1. Download and install Java SE Development Kit (JDK) version 17. You can download it from either of the following sources:

    Note: After installation, remember to set the JAVA_HOME environment variable to the directory where JDK was installed.

  2. Download and install Ballerina Swan Lake.

  3. Download and install Docker.

    Note: Ensure that the Docker daemon is running before executing any tests.

  4. Export Github Personal access token with read package permissions as follows,

    export packageUser=<Username>
export packagePAT=<Personal access token>

Build options

Execute the commands below to build from the source.

  1. To build the package:

    ./gradlew clean build

  2. To run the tests:

    ./gradlew clean test

  3. To build the without the tests:

    ./gradlew clean build -x test

  4. To run tests against different environments:

    ./gradlew clean test -Pgroups=<Comma separated groups/test cases>

  5. To debug the package with a remote debugger:

    ./gradlew clean build -Pdebug=<port>

  6. To debug with the Ballerina language:

    ./gradlew clean build -PbalJavaDebug=<port>

  7. Publish the generated artifacts to the local Ballerina Central repository:

    ./gradlew clean build -PpublishToLocalCentral=true

  8. Publish the generated artifacts to the Ballerina Central repository:

    ./gradlew clean build -PpublishToCentral=true

Contribute to Ballerina

As an open-source project, Ballerina welcomes contributions from the community.

For more information, go to the contribution guidelines.

Code of conduct

All the contributors are encouraged to read the Ballerina Code of Conduct.

Useful links