Perpetual

# The Perpetual Project

LLM-driven software development assistant.

## Description

`Perpetual` is an LLM-driven software development assistant designed to enhance the productivity and efficiency of software developers. Its primary function is to streamline the coding process by automating the generation and modification of code based on textual descriptions provided by programmers. It achieves this by analyzing the project's codebase and interpreting programmer instructions embedded as special comments or by reading tasks directly from text files or standard input. `Perpetual` can generate new code or entities in existing or new files, make changes to existing code, write or refine documentation, explain the code and provide answers according to queries.

The program works from the command-line and does not require complex installation, frontend/backend or containerization. It focuses on direct interaction with the project's codebase, eliminating the need for additional tools, deployment, or server infrastructure (apart from LLM API access keys). This approach results in a simple and easily deployable tool that can be used directly by developers or integrated into larger AI software development ecosystems. Main operations of the program support reading input from stdin and writing easily-parsable output to stdout (with redirecting logs to stderr). So it can be easily used with multi-agent systems as a main tool for writing code and planning project architecture.

The program operates strictly inside the user's project directory and it only allowed to modify source code files inside it, ensuring a controlled and safe environment for code manipulation.

**TL;DR: Go straight to Examples**:

- [Tetris](docs/example_2025_05/tetris.md)
- [Conway's Game of Life](docs/example_2025_03/game_of_life.md)
- [Fractal Visualizer](docs/example_2025_02/fractal_visualizer.md)

## Limitations

[See current technical limitations](docs/limitations.md)

### Warning

While `Perpetual` t