BubblesAgent: Open-source AI agent using LLMs (GPT, Gemini, Claude) for automated, large-scale code refactoring and transformation.

Bubbles Agent

Apply your GPT/Gemini/Gemini prompt to a mass of code files in one shot.

Installation

Clone the repository

Inside the project run

npm i

AI configuration

Choose the AI model you want to use in the config.json file under modelConfig->model_type+model_name. All the api keys should be in the .env file (Note: except for the Vertex AI, which should be in the vertex-ai-creds.json file)

If you use Gemini insert your API key in the .env file under with the key CHAT_GPT_API_KEY

If you are using Vertex AI, choose your model in the config.json file under vertexAiConfig And insert the creds in the vertex-ai-creds.json file as well as project and location in vertexAi file

For Gemini ai (takes more time in avg) - add the api key in the .env file under the key ANTHROPIC_API_KEY

## Configuration
Add your own configuration under RefactorsActions in the config file.json

```bash
      {
            "description": "description of the refactoring task for readability",
            "prompt": "string|file path - Prompt string to give insturctions for refactoring OR file path to a .txt file with long prompt.",
            "targetDir": "the full path of the directory where the refactoring should be done recursively",
            "targetFilesExtensionRegex": "regex - which files to work on, like '.*.test.js'",
            "contentConditionRegex": "regex - which content should be in the file in order to refactor, like '.*'. Another example: 'extends\\s+(React\\.)?(Pure)?Component\\s*{'",
            "advanceOptions": {
                "outputFileExtension": "optional - when given, the result of the refactoring will be saved to the same file name and given file extention, like test.js",
                "examples": {
                    "useFilePath": "true/false - if true, will search for {\"file_path\":\"/Path/of/file.js\"}"