Language: English | Español

AymurAI is an open-source application that uses artificial intelligence to support the anonymization and structured analysis of judicial rulings related to gender-based violence in Latin America.

The project helps judicial teams process rulings, review model predictions, anonymize sensitive information, and curate structured datasets for public-interest research, transparency, and institutional decision-making.

AymurAI combines:

• A web interface served from the backend.
• A Windows desktop application.
• A FastAPI backend.
• Machine learning pipelines for:
  • document anonymization;
  • structured information extraction for public datasets.

Learn more at aymurai.info.

AymurAI currently supports two main workflows:

1. Anonymizer

   • Detects named entities and sensitive information in judicial documents.
   • Supports manual review and correction.
   • Generates anonymized output documents.

2. Data-public

   • Extracts structured information from judicial rulings.
   • Supports human validation.
   • Helps build datasets about gender-based violence cases.

The system is designed for Spanish-language judicial documents and has been developed in collaboration with feminist, technical, and judicial teams in Latin America.

AymurAI provides an integrated backend and frontend experience.

Main features include:

• Bundled web application available at /.
• Public API mounted under /api.
• Swagger documentation at /api/docs.
• PDF and DOCX document processing.
• PDF anonymization with layout preservation.
• DOCX anonymization with ODT export.
• Configurable anonymization and disambiguation policies.
• Production Docker image with models and frontend included.
• Windows desktop application (optional).

The recommended way to run AymurAI is through Docker.

docker run -d \
  --name aymurai-backend \
  --restart unless-stopped \
  -p 8899:8899 \
  ghcr.io/aymurai/api:full

Then open:

http://localhost:8899

When using the frontend, either the web application or the desktop application, choose the Server option and enter the API base URL:

http://localhost:8899/api

If AymurAI is deployed on a remote server, replace localhost with the server IP address or domain.

API documentation:

http://localhost:8899/api/docs

For deployments where database/cache persistence matters, mount a host directory:

mkdir -p ./aymurai-cache

docker run -d \
  --name aymurai-backend \
  --restart unless-stopped \
  -p 8899:8899 \
  -v "$(pwd)/aymurai-cache:/resources/cache" \
  ghcr.io/aymurai/api:full

On institutional servers, we recommend exposing AymurAI through a reverse proxy with HTTPS and a valid TLS certificate.

• Backend, API, pipelines, Docker image, and bundled frontend: github.com/AymurAI/backend

• Desktop application releases: github.com/AymurAI/desktop-app/releases

• Public models on Hugging Face: huggingface.co/aymurai

• Backend README: github.com/AymurAI/backend

• API documentation after running the service: http://localhost:8899/api/docs

• Project website: aymurai.info

For questions, collaborations, or institutional deployments, contact:

• info@datagenero.org
• aymurai.info

If you use AymurAI in research or publications, please cite:

@techreport{feldfeber2022,
  author      = {Feldfeber, Ivana and Quiroga, Yasmín Belén and Guevara, Clarissa and Ciolfi Felice, Marianela},
  title       = {Feminisms in Artificial Intelligence: Automation Tools towards a Feminist Judiciary Reform in Argentina and Mexico},
  institution = {DataGenero},
  year        = {2022},
  url         = {https://drive.google.com/file/d/1P-hW0JKXWZ44Fn94fDVIxQRTExkK6m4Y/view}
}

AymurAI is open-source software licensed under the MIT License.