{"categories":[{"label":"Artificial Intelligence","url":"https://skillfed.io/packages/category/scientific-engineering-artificial-intelligence/2"}],"enrichment":{"capability":"SWE-smith generates large-scale software engineering training datasets by synthesizing task instances from GitHub repositories and managing their execution environments via Docker.","skillfed_tags":["code-generation","benchmark-dataset","agent-training"],"use_cases":["Generate custom software engineering benchmarks from any GitHub repository by creating execution environments and synthesizing task instances.","Fine-tune language models on software engineering tasks using the pre-built dataset of 52k task instances.","Create reproducible training data for software engineering agents with controlled task generation and test-based filtering.","Build reinforcement learning datasets for code-based models using the container infrastructure.","Evaluate software engineering agents on diverse repositories by synthesizing and executing task instances in isolated Docker environments."],"what_it_does":"SWE-smith is a toolkit for generating software engineering training datasets at scale. It converts GitHub repositories into executable environments and synthesizes task instances\u2014such as file localization or program repair tasks\u2014that can be used to train language models to perform software engineering work. The package manages Docker-based execution environments, generates synthetic tasks, filters them by test coverage, and produces issue descriptions for training.\n\nThe toolkit is designed for researchers and practitioners building datasets for software engineering agents. Installation from source is required to build custom datasets; the package depends on Docker and is tested only on Ubuntu 22.04.4 LTS, with explicit non-support for Windows and macOS.","worth_installing":"Yes, if you are training software engineering agents or building custom code benchmarks and have Docker and Ubuntu 22.04.4 LTS available. The toolkit is actively maintained, has low install friction, and provides a complete pipeline from repository to training dataset. Not suitable for Windows or macOS users, and requires significant infrastructure to use beyond pre-built resources."},"id":"swesmith","links":{"html":"https://skillfed.io/packages/swesmith","md":"https://skillfed.io/packages/swesmith.md","pypi":"https://pypi.org/project/swesmith/"},"maintenance":{"status":"active"},"meta":{"latest_release":"2026-02-27","license_spdx":null,"license_treatment":"permissive","name":"swesmith","python_support":"supports_current","summary":"The official SWE-smith package - A toolkit for generating software engineering training data at scale."},"popularity":{"monthly_downloads":4979918,"position":2188,"tier":"top_5000"},"security":{"n_vulnerabilities":0},"version":"0.0.9"}