AymurAI

Language: English | Español

AymurAI provides an integrated application, API, and ML pipelines for processing judicial rulings through two main workflows:

• anonymizer: extract named entities and produce anonymized documents.
• data-public: extract structured information for public dataset curation.

This repository contains the React/Electron frontend, FastAPI service, production pipeline configurations, and database persistence used by both workflows.

About AymurAI

AymurAI is a project focused on supporting the generation of anonymized and structured judicial data for gender-based violence (GBV) cases in Latin America. The application guides users through document ingestion, ML inference, manual review, validation, and export for operational and research uses.

The frontend can run as a web application served by FastAPI or as an Electron desktop application. It supports document upload and preview, interactive annotation review, entity policy configuration, anonymized document export, and structured dataset validation.

Documentation

• Technical docs index: docs/README.md
• API reference: docs/api/README.md
• Pipelines index: docs/pipelines/README.md
• Anonymizer flow: docs/pipelines/anonymizer/README.md
• Datapublic flow: docs/pipelines/datapublic/README.md
• Internal database schema: docs/database/README.md
• Frontend development and packaging: frontend/README.md

Quick Start (Docker Image)

Run the full API image (includes production resources):

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

Optional: persist DB/cache outside the container (host volume mounted at /resources/cache):

mkdir -p ./aymurai-cache

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

Optional: GPU runtime (requires NVIDIA Container Toolkit):

docker run -d --name aymurai-backend-gpu --gpus all \
  -e TORCH_DEVICE=cuda \
  -p 8899:8899 \
  ghcr.io/aymurai/api:full

Open Swagger UI:

http://localhost:8899/api/docs

Open the bundled frontend:

http://localhost:8899/

Quick Start (Docker Compose)

Use the services defined in docker-compose.yml:

# CPU, lightweight API profile
make api-up

# CPU, full API profile
make api-full-up

# GPU, lightweight API profile
API_SERVICE=aymurai-api-gpu make api-up

# GPU, full API profile
API_FULL_SERVICE=aymurai-api-full-gpu make api-full-up

Check logs:

make api-logs
# or make api-full-logs

Runtime Overview

• Framework: FastAPI
• Default API port: 8899
• Bundled frontend path: GET /
• Frontend implementations: browser and Electron
• DB engine: SQLModel + Alembic migrations on startup
• Default DB URI: sqlite:////resources/cache/sqlite/database.db
• Production pipeline configs:
  • resources/pipelines/production/flair-anonymizer/pipeline.json
  • resources/pipelines/production/datapublic/pipeline.json

Main Public Endpoints

• GET /api/server/healthcheck
• GET /api/server/stats/summary
• POST /api/misc/document-extract (and deprecated alias POST /api/document-extract)
• POST /api/anonymizer/predict
• POST /api/anonymizer/disambiguate
• POST /api/anonymizer/validation
• POST /api/anonymizer/anonymize-document
• POST /api/datapublic/predict/{document_id}
• GET /api/datapublic/validation/document/{document_id}
• POST /api/datapublic/validation/document/{document_id}

For full request/response contracts and examples, see docs/api/README.md.

Closed-Network Deployment

To move an image into a closed environment:

docker image save ghcr.io/aymurai/api:full -o aymurai-api-full.tar
docker load -i aymurai-api-full.tar

Contributing

Contributions are welcome across documentation, API, and pipeline improvements.

• Contributing guide: docs/CONTRIBUTING.md
• Security and ethics: docs/SECURITY.md
• Code of conduct: docs/CODE_OF_CONDUCT.md

Contributors

• Julián Ansaldo - @jansaldo at collective.ai (email)
• Raúl Barriga - @jedzill4 at collective.ai (email)
• Sofía del Pozo - @sofiadelpozo at collective.ai (email)
• Paolo Donizetti - @padonizetti at collective.ai (email)
• Conrado Beatriz - @conrabeatriz at collective.ai (email)
• Lionel Chamorro - @lionelchamorro at collective.ai (email)
• Damián Mazzini - @ dmazzini at collective.ai (email)

Citing AymurAI

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

@techreport{feldfeber2022,
  author      = {Feldfeber, Ivana and Quiroga, Yasm\'{\i}n Bel\'{e}n 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}
}

@inproceedings{10.1145/3706598.3713681,
  author    = {Ciolfi Felice, Marianela and Feldfeber, Ivana and Glasserman Apicella, Carolina and Quiroga, Yasm\'{\i}n Bel\'{e}n and Ansaldo, Juli\'{a}n and Lapenna, Luciano and Bezchinsky, Santiago and Barriga Rubio, Ra\'{u}l and Garc\'{\i}a, Mail\'{e}n},
  title     = {Doing the Feminist Work in AI: Reflections from an AI Project in Latin America},
  booktitle = {Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems},
  series    = {CHI '25},
  year      = {2025},
  isbn      = {9798400713941},
  publisher = {Association for Computing Machinery},
  address   = {New York, NY, USA},
  doi       = {10.1145/3706598.3713681},
  url       = {https://doi.org/10.1145/3706598.3713681},
  abstract  = {The contemporary AI development landscape is dominated by big corporations, lacks diversity, and mostly centres the Global North, or applies extractivist logics in the South. This paper showcases a feminist process of AI development from Latin America, where we created an interactive, AI-powered tool that helps criminal court officers open justice data, addressing a data gap on gender-based violence. Through a collaborative autoethnography, drawing from Latin American feminisms, we unpack and visibilize the feminist work that was required, as a crucial step to counter hegemonic narratives. Foregrounding the subjugated knowledges of our experiences, we offer a concrete example of a feminist approach to AI development grounded in practice. With this, we aim to critically inspire those who consider building technology in service of social justice causes, or who choose to build AI systems otherwise.},
  articleno = {998},
  numpages  = {18},
  keywords  = {Global South, NGO, activism, critical HCI, critical computing, duoethnography, feminist AI, feminist research},
  location  = {},
}

For the most up-to-date citation, please use the project-level reference in the organization repository: github.com/aymurai.

License

AymurAI is open-source software licensed under the MIT License. This license allows modification, distribution, and private use, provided that appropriate credit is given to the original authors.