mini-devin-gemini

A Devin-like tool powered by Gemini, designed for enhanced code generation and task management.

Features

Step-by-step software creation: Breaks down complex projects into manageable tasks.
AI-powered task management: Uses Gemini to generate, execute, and track tasks.
Automated code generation: Creates code files based on task descriptions.
Command execution: Executes shell and Node.js commands.
File manipulation: Supports writing, appending, reading, listing, and deleting files.
Child process management: Spawns, monitors, and terminates child processes.
Session memory: Retains conversation history for context and continuity.
SQLite integration: Stores tasks and session data for persistence.
Error handling and retries: Handles command errors with retries and AI-guided recovery.
Detailed logging: Logs LLM interactions and task outputs for review.

Set up your environment:
- Create a .env file in the project root.
- Add your Gemini API key: GEMINI_API_KEY=YOUR_API_KEY
Run the application:
```
node index.js
```
Follow the prompts:
- You'll be asked to describe the project you want to build.
- Specify the working directory where the project files will be created.

The AI will then generate a plan and start executing tasks. Follow the console output for progress and any additional instructions.

Gemini Can run the following command for achieving the task you want

Supported Command Types:
- shell: Executes shell commands.
- node: Executes Node.js commands.
- write_file: Creates or overwrites files with specified content.
- append_file: Appends content to existing files.
- spawn_child_process: Starts a child process.
- retrieve_child_process: Retrieves logs from a running child process.
- close_child_process: Terminates a child process.
- list_files: Lists files in the working directory.
- read_file: Reads the content of a file.
- delete_file: Deletes a file.

The code currently is in not structured mode, any commit to restrutcture or refactor code is welcome!

If you want to change the model, you can change the model inside the index.js at Line around 742

Improved code analysis: More robust error detection and correction.
Support for more languages: Expand beyond shell and Node.js.
Plugin system: Allow users to extend functionality.
Interactive debugging: Enable interactive debugging sessions with the AI.
Pinecone Integration: Enable integration with Pinecone for effective super-duper long context length.