Genaios

TextMachina: Build Datasets for Machine-Generated Text Tasks

TextMachina is a modular Python framework for creating datasets for machine-generated text detection, attribution, boundary detection, and mixcase tasks. It connects to configurable LLM providers and includes tools for prompting, bias mitigation, and dataset exploration.

View on GitHub Compare
Stars
21
Forks
0
Stars, last 30 days
+0
Last commit
7 months ago
Contributors
3

Our take

Quietno commits in the last 90 days and no release in a year

Health score

23/100

How it is scored
Activity
19
Community
30
Issues
21
Pull requests
20

+0 stars in the last 30 days

  • Fewer contributors than 82% of the projects we track

TextMachina brings several MGT dataset-building tasks and quality checks into one configurable Python framework, but its quiet maintenance and concentrated contributor activity warrant caution for long-term adoption.

Good fit if

  • You need datasets for one or more of its supported MGT tasks and can configure a supported LLM provider.
  • Your team is comfortable using Python and wants both CLI and programmatic workflows.
  • You can review generated samples for artifacts and biases before relying on a dataset.

Look elsewhere if

  • You need a project with recent release activity or evidence of active issue and pull-request handling.
  • You need well-supported multilingual or machine-generated code datasets, which the project identifies as limitations.
  • Your intended use requires commercial rights or permission to modify and redistribute the project materials without first clarifying the license terms.
All health signals
Last commit2026-03-07 (7 months ago)
Commits, last 90 days0
Releases, last 12 months0 (latest 0.2.12, 2024-05-08)
Contributors3 (top contributor: 78% of commits)
Issues closed, last 90 days0
Pull requests merged, last 90 days0
Project age2 years

Checked on 2026-10-16 with the GitHub API.

Project history

Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.

Selected snapshot

Stars

21

Forks

0

Change since Nov 17, 2026: 0 stars·0 forks

Overview

TextMachina is a modular Python framework for building datasets for machine-generated text (MGT) tasks. It connects to configurable LLM providers, helping researchers and developers create data for detecting generated text, identifying its source model, finding human/generated boundaries, and labeling mixed human and generated passages.

The framework addresses the work of preparing prompts, generating examples, and checking dataset quality. It offers both a CLI and a Python API, with configurable extractors, decoding constraints, post-processing, and metrics. Its intended users are teams developing or evaluating MGT-related models who need to assemble task-specific datasets.

Key Features

  • Generate datasets for MGT detection, model attribution, boundary detection, and mixcase tagging.
  • Configure models from providers including OpenAI, Anthropic, Cohere, Google Vertex AI, Amazon Bedrock, Azure OpenAI, Hugging Face, and inference servers.
  • Use extractors to fill prompt templates with information from human text.
  • Apply decoding constraints inferred from human text and post-processing intended to reduce artifacts and biases.
  • Explore small generated samples interactively and inspect dataset metrics before full generation.
  • Run generation through the CLI or Python API, with cached CLI runs that can resume after interruption.

Use Cases

  • Researchers preparing labeled human and LLM-generated examples for detection-model training or evaluation.
  • Teams comparing attribution models that identify which LLM generated a passage.
  • Dataset builders creating examples with human-to-generated transitions for boundary detection.
  • NLP practitioners constructing mixed human and generated text for span- or sentence-level classification.

What you need

Detected in the repository

  • Python (from setup.py)
  • Automated checks on GitHub Actions

Getting Started

Install the package and optional provider dependencies:

pip install 'text-machina[all]'

See the README for configuration examples and CLI and Python API usage.

License in Plain Words

No standard license was detected. Without a license, the default copyright rules apply: you may read the code, but not reuse it. Check the repository or ask the authors.

Considerations

The project is marked quiet: it has had no commits in the last 90 days and no release in over a year. It has three contributors, with most of the code coming from one person, and there have been no recent issue closures or merged pull requests in the measured period. The repository data does not assert a license, while the README describes CC BY-NC-ND 4.0 terms that restrict commercial use and distribution of modified material; verify the applicable terms before adopting or redistributing it. The README also notes limited multilingual support, limited support for generated-code datasets, and the need to inspect outputs for disclosure patterns and other biases. The repository requirements data indicates Python and CI, but no tests are reported.

Found this useful?

Share it with someone who would like TextMachina.