prmpt

prmpt is a Rust utility for turning an entire code repository into a prompt that a Large Language Model can understand. It also reinjects the model's answers back into the same repository. Use prmpt when you want a repeatable way to ship code in and out of an LLM without manually copying files.

Why prmpt?

• One command prompt generation – walk your project, honor .gitignore, and produce a single text file ready for an LLM.
• Docs‐only extraction – gather just the docstrings or comments so the model focuses on high level documentation.
• Jupyter notebook support – include cell outputs when you need them.
• Safe injection – prmpt reads the LLM output and places each code block into the correct file.

Install

1. Install a Rust toolchain.
2. Build prmpt from source:

git clone https://github.com/labiium/prmpt.git
cd prmpt
cargo build --release

The executable will be at target/release/prmpt.

Or

cargo install --git https://github.com/labiium/prmpt.git

and the prmpt command will then be available.

Command overview

prmpt exposes two main subcommands. You can also invoke named configurations stored in a prmpt.yaml file.

Generate

Create a prompt file from a repository.

• Basic run

prmpt generate --path my_project --language rust --output repo.out

• Important flags
  • --ignore <pattern> – repeat to skip files or directories.
  • --docs-comments-only – extract docstrings and comments without source code.
  • --delimiter <token> – fence used around each block (defaults to ```).

Inject

Read an LLM's output and write changes back into the repository.

• Basic run

prmpt inject --input llm_output.txt --path my_project

The input file should contain a path followed by a fenced code block for every change:

    src/main.rs
    ```rust
    fn main() {
        println!("hello updated world");
    }
    ```

Running with a configuration

A prmpt.yaml file allows named setups. A minimal example is below.

base:
  path: ./my_project
  language: rust
  output: prompt.out
  ignore:
    - target
  prompts:
    - "Summarise the project before rewriting."

Execute that configuration simply by running:

prmpt base

Example output

Running prmpt generate on a small project might produce something like:

    sample_project_1
    ├── README.md
    └── main.py

    ```README.md
    # Sample Project 1
    ```

    ```main.py
    print("Hello, world!")
    ```

This output can be fed directly into your favourite LLM. After editing, feed the modified file back to prmpt inject.

Library usage

prmpt can also be used as a library. The Generator and Injector types expose the same logic as the CLI:

use prmpt::{Config, Generator, Injector, run_and_write};

let config = Config { path: Some("my_project".into()), ..Default::default() };
let generator = Generator::default();
run_and_write(&generator, &config)?;

prmpt streamlines the tedious parts of preparing code for an LLM and applying the results. Give it a try on your next project and focus on the creative parts instead of file management.