Vibe Code Suite

Vibe Code Suite

===================

**Vibe Code Suite is a python program that slices out functions with the ast library provided by python and sends them to Gemini to surgically refractor code based on the users prompt. It gives the sliced function to Gemini saving tokens and prevent the AI from forgetting the context of your code.**

## Why Vibe?

* **AST-Powered Context:** Surgical slicing ensures the LLM sees only the logic you care about.

* **Token Efficiency:** Reduces overhead by up to 80% on large files.

* **Precision Refactoring:** Built-in support for **GPT-OSS-120B** via Groq for elite Python optimizations.

* **Hacker-First UI:** A sleek, Tokyo Night-themed TUI built with **Textual**.

## Requirements

* **Python 3.9+** (Required for ast.unparse)

* **Libraries:** pip install textual requests

* **API Access:** A valid Google Gemini API Key.

## Technical Stack

- **Engine:** Python 3.12+ (leveraging ast.unparse)

- **Intelligence:** GPT-OSS-120B (Groq API)

- **Interface:** Textual TUI

- **Optimized For:** Low-end Android tablets (Like a Lenovo YT-X705X). Anything that can run Python without crashing.

## Usage

1. Start main.py with python3 main.py (assuming python is installed).

2. Select the file from the directory tree (file has to be in a subdirectory or same directory as the main.py).

3. Enter the prompt, focus function (the function to slice in the code (Classes aren't supported but functions inside classes are supported)) and your Gemini API key.

4. Click Vibe!

It will save the file to ai-output.py.

Vibe Code Suite's primary innovation is the AST Slicing Engine, which reduces noise and token cost for any LLM. While the demo uses the Groq API for accessibility on mobile hardware, the engine is designed to be provider-agnostic. It can easily be configured to connect to a local Ollama or vLLM instance for users who require a 100% offline, FOSS-only workflow. (Do this at your own risk.)