FabRAG (FabAcademy/Digital Fabrication/Fabulous Retrieval-Augmented Generation) is a project born at Fablab Leon, during the the global FabAcademy instructors' bootcamp.
The goal of this project to enhancing access to and interaction with the wealth of knowledge contained within the Fab Academy curriculum using Artificial Inteligence. FabRAG seeks to create an intelligent system capable of understanding, retrieving, and generating relevant information from the Fab Academy's extensive educational resources.
- To create comprehensive datasets based on Fab Academy content
- To develop an AI-powered system capable of understanding and responding to complex queries about digital fabrication, assisting new students and helping them leaverage the knowledge of the Fab Academy community
- To preserve and respect the intellectual property rights of content creators
- To provide an intuitive interface for students, educators, and makers to access Fab Academy knowledge
Inspired by the Fab 1.0 to 5.0 analogy, the FabRAG project is divided into five main phases, each building upon the previous to create an increasingly sophisticated and capable system.
In this initial phase, we focus on creating a foundational Retrieval-Augmented Generation (RAG) system using the text extracted from selected Fab Academy web pages. This phase establishes the core functionality of information retrieval and response generation along with the distribution mechanism for the system.
This phase is already completed. You can test the instructions and data required to set up the system in the FabRAG_1 repository.
Building on Phase 1, we incorporate additional metadata into our system, including authorship information, content's licensing and source URLs. This enhancement allows for more accurate attribution and enables users to trace information back to its original source.
In this phase, we extend the system to support multimodal search, incorporating image analysis alongside text-based search. This feature will allow users to find information based on visual content as well as textual queries, enhancing the system's versatility.
Phase 4 focuses on improving the system's understanding and response capabilities. We distill thematic questions from the most relevant content and use these to fine-tune the base language model, resulting in more accurate and context-aware responses.
In the final phase, we implement intelligent agents capable of extracting new information from linked content within the current year's Fab Academy materials. This dynamic approach ensures that the system stays up-to-date with the latest developments and provides users with the most current and relevant information.
Throughout all phases of the project, we maintain a strong commitment to ethical considerations, including:
- Respecting intellectual property rights and adhering to copyright laws
- Ensuring proper attribution for all content
- Maintaining transparency about the AI-generated nature of responses
- Protecting user privacy and data security
As the FabRAG project evolves, we anticipate exploring additional enhancements such as:
- Multi-language support to make Fab Academy content accessible to a global audience
- Collaborative features to facilitate knowledge sharing among users
We welcome contributions from the Fab Academy community and beyond. Whether you're a digital fabrication expert, a machine learning enthusiast, or a user experience designer, there are many ways to get involved and help shape the future of FabRAG.
This project builds upon previous work to highlight the existing talent in the FabAcademy network. In particular, the first version of this dataset builds upon the Expert Network Map created by Adam Stone and the weekly posts he produced highlighting the most referenced content in the Fab Academy curriculum. Thanks Adam for your dedication to this project!
For additional information about the project, check the interactive expert network map and/or the documentation about the project.
This project development will be supported by several Generative AI models over time. In particular, the development started to test the new features of Claude Sonnet model and the projects feature, but this might change over time.
[Appropriate license information to be added here]
FabRAG: fabrag@lahoramaker.com
FabRAG is an ongoing project, and this README will be updated as the project progresses through its various phases. Stay tuned for more exciting developments!