Urbanite is a framework for human-AI collaboration in urban visual analytics that leverages a dataflow-based model allowing users to specify intent at multiple scopes, enabling interactive alignment across specification, process, and evaluation stages. The framework incorporates features for explainability, multi-resolution task definition across dataflows, nodes, and parameters, and supporting interaction provenance based on findings from a survey identifying existing challenges.
Urbanite: A Dataflow-Based Framework for Human-AI Interactive Alignment in Urban Visual Analytics
Gustavo Moreira, Leonardo Ferreira, Carolina Veiga, Maryam Hosseini, and Fabio Miranda
Paper: Under Review
urbanite_compressed.mp4
This project is part of the Urban Toolkit ecosystem, which includes Urbanite, Curio and UTK.
- Provenance-aware dataflow
- Modularized and collaborative visual analytics
- Support for 2D and 3D maps
- Integration with UTK and Vega-Lite
- LLM-based features:
- Dataflow generation and scaffolding
- Task & subtask definition
- Code generation
- Connection suggestions
- Dataflow- or node-level explanations
- Provenance and data inspection
For detailed instructions on how to use the project, please see the usage document. A set of examples can be found here.
🚀 Urbanite now supports a Docker-based setup for easier installation and orchestration of all components. See the usage guide for instructions on running Urbanite with Docker.
If you'd like to contribute, see the contributions document for guidelines. For questions, join UTK's Discord server.
Gustavo Moreira (UIC)
Leonardo Ferreira (UIC)
Carolina Veiga (UIUC)
Maryam Hosseini (UC Berkeley)
Fabio Miranda (UIC)
Urbanite is MIT Licensed. Free for both commercial and research use.