DataChat is multilingual Open-Source natural language interface for Dataverse and other data platforms with experimental Graph AI implementation for Croissant ML support. DataChat can literally talk back to you and explain what is inside of every single dataset, you can ask any questions and it's responding on the level of metadata described by Croissant standard.
This AI Agent is being developed by Slava Tykhonov from DANS-KNAW (Data Archiving and Networked Services, the Netherlands).
DataChat is in active development, please find below the acknowledgements for resources and contributions from the ongoing projects.
| Region | Project | Funding information | Component |
|---|---|---|---|
| Netherlands | ODISSEI | NWO grant number 184.035.014 | ODISSEI Portal |
| Netherlands | SSHOC.nl | NWO grant number 184.036.020 | SSHOC.nl data platform |
| France | Now.Museum | Université Paris Cité microgrant | Now.Museum Timeline |
| European Union | MuseIT | HORIZON-CL2-2021-HERITAGE-01-04, Grant agreement #101061441 | AI for people with disabilities |
cp env_sample .env
docker-compose up -d
For local deployment of llama3:
docker exec -it ollama /bin/bash
ollama pull llama3
Demo of the app below:
For academic use please cite this work as:
Tykhonov, Vyacheslav. (2024). Building natural language interface for Dataverse network based on Croissant ML standard. Zenodo. https://doi.org/10.5281/zenodo.13842869