miso-belica/sumy

Sumy is a Python library and command-line tool that automatically condenses long text documents or web pages down to a shorter summary. You point it at a URL or a plain text file, tell it how many sentences you want, and it returns a condensed version of the content. The summarization is done without sending your data to any external AI service, so everything runs locally on your machine. The library includes several different summarization methods, each based on a different mathematical approach to deciding which sentences carry the most meaning. The README links to documentation that explains each method. You can try them out directly from the command line, for instance by running a single command against any Wikipedia article to get a ten-sentence summary back immediately. Support for different languages is a built-in feature. The README notes there is a good chance your language is already supported, and it also explains how to add a new one if not. This makes sumy useful for people working with text in languages other than English. If you want to use sumy inside your own Python project rather than from the command line, the library exposes a straightforward API. You pick a parser (HTML or plain text), choose a summarizer, set the sentence count, and iterate over the results. The README includes a short working code sample that shows exactly how to wire these pieces together. The package also includes an evaluation framework that lets you compare a generated summary against a reference summary you provide. This is useful if you are testing whether a particular method works well for your content type. Installation is done through pip or uv, and a Docker image is available if you prefer not to install anything locally.