Release v0.1

For our first assignment in the open source class, we were tasked to make a release 0.1 for a CLI tool that utilizes LLMs. It can be any kind of tool as long as it fulfills a set of requirements: It has to be using OpenAI's chat completion and it has to be about processing and transforming files through command line arguments.

Genereadme

For my assignment, I chose to create Genereadme, which is a CLI tool that generates a README file for the source code files provided. I chose to do this project as writing a README documentation is not one of my strong suits, to which I also find a hassle. But considering how important documentations are, especially in big projects, this step in the development process cannot be ignored. So I figured, why not make something that could help me do this instead?

GROQ_API_KEY=API_KEY

or

genereadme -a API_KEY
genereadme --api-key API_KEY

npm install

genereadme <files>
genereadme examples/sum.js
genereadme examples/createUser.js examples/sum.js

Get started

Simply install the packages after cloning the repository

npm install

Run and generate!

genereadme examples/sum.js
genereadme examples/createUser.js
genereadme examples/sum.js examples/createUser.js
genereadme examples/sum.js examples/createUser.js --output sample.md

How it works

The user can pass any number of files in the command line using the command genereadme <file1> <file2> .... The contents of these files will be processed individually to generate their README documentation.
The generated READMEs will be saved as filename_README.md to prevent any naming collision when processing multiple files.

Arguments

• <filename>: The file to be processed and generate a README for.

Options

• -v or --version: Displays the tool's name and the current version.
• -h or --help: Displays the help message for the tool.
• -o or --output: Outputs the generated results into the specified output filename.
• -a or --api-key: Allows the user to use their own valid API key.
• -t or --temperature: Allows the user to set their preferred temperature for the chat completion.
• -tu or --token-usage: Shows the amount of tokens used for the prompt and the completion.

Examples

genereadme examples/createUser.js

export async function createUser(data) {
    const user = {
      Username: data.email,
      UserPoolId: process.env.AWS_COGNITO_POOL_ID,
      TemporaryPassword: data.temporaryPassword,
      UserAttributes: [
        {
          Name: "email",
          Value: data.email,
        },
        {
          Name: "name",
          Value: data.name,
        },
      ],
      MessageAction: MessageActionType.SUPPRESS,
      DesiredDeliveryMediums: [DeliveryMediumType.EMAIL],
    };

    const command = new AdminCreateUserCommand(user);
    try {
      const createRes = await cognitoClient.send(command);
      logger.info(`Created user: [${JSON.stringify(createRes)}]`);

      const addUserToGroupParams = {
        UserPoolId: process.env.AWS_COGNITO_POOL_ID,
        Username: data.email,
        GroupName: data.group,
      };

      const addUserToGroupCommand = new AdminAddUserToGroupCommand(addUserToGroupParams);
      const addRes = await cognitoClient.send(addUserToGroupCommand);
      logger.info(`Added user to group: [${JSON.stringify(addRes)}]`);
    } catch (error) {
      logger.error(`Error creating user: ${error}`);
      throw new Error("Error creating user.");
    }
  }

generated README for createUser.js

Conclusion

It is my first time working directly with LLMs by myself, so it definitely took some time prompt engineering to get a somewhat satisfying result. However, I do know that there are still a lot that can be improved on and I'm already getting ideas on what to do and what else I can add to this project!