NitroDigest

TL;DR your data, privately.

NitroDigest – the privacy‑first, local‑LLM text‑summariser for developers.

This project is in alpha phase.

Features

• Runs 100 % on‑device with Ollama – your private data never leaves localhost
• Command-line interface with various options
• Completely free (open source, MIT license)

Usage

Prerequisites

To run this tool, you needs to have Ollama and Python installed on your local machine.

Installation

pip install nitrodigest-cli

Basic Usage

Run NitroDigest with the default configuration:

nitrodigest <file or directory you want to summarize> > <destination where to want to save summary>

Examples

Process current directory (summarize all files in the current working directory):

nitrodigest > summary.md

Summarize one file and save it to summary.md:

nitrodigest my_long_text.html > summary.md

Summarize files in a directory and save them in a summary.md:

nitrodigest my_directory/ > summary.md

Command Line Arguments

You can override any configuration setting using command line arguments:

nitrodigest \
    --model mistral \
    --timeout 800

Available arguments:

• --timeout: Time in seconds for API requests to Ollama (default: 300)
• --prompt-file: Path to custom prompt template file (overrides default one)
• --prompt: Direct prompt content (overrides prompt-file)
• --model: Model that will be used for summarization (default: mistral)
• --ollama_api_url: URL of Ollama API (default: http://localhost:11434)
• --format: Output format. Can be text or json (default: text)
• --include-original: Include original text in the summary output (default: False)
• --max-workers: Maximum number of parallel workers for directory processing (default: 4)

Custom Prompt Configuration

You can specify a custom prompt in two ways

1. Using the --prompt-file argument:

nitrodigest --prompt-file custom_prompt.txt

2. Passing the prompt content directly:

nitrodigest --prompt "$(cat my_awesome_prompt.txt)"

The prompt template should contain placeholders:

• {metadata}: it's used to render information like source, title, and date
• {text}: For the text content to be summarized

Contributing

Do you want to contribute this tool? Check the Contributing page:

Getting started

Report an issue

Found an issue? You can easily report it here:

https://github.com/Frodigo/garage/issues/new

License

This project is licensed under the MIT License - see the LICENSE file for details.