LearningLion - Kamervragen
Welcome to the git repo of the use case 'Kamervragen'. Here, we document our progress regarding our feasibility study into RAG for supporting the process of answering questions from members of parliament.
Description of the project
The project is a study on the use of generative AI to improve the services of SSC-ICT by supporting employees and optimizing internal processes. In particular, the focus is on generative large language models (LLM), in the form of Retrieval Augmented Generation (RAG), because they can have the most significant impact on the daily work of SSC-ICT employees. The assignment (Speech Recognition & AI) has been approved by the SpB and has been running since the beginning of 2023.
The original Learning Lion Repository
The Learning Lion repository is based on the repository made by PBL under the direction of Stefan Troost (https://github.com/pbl-nl/appl-docchat), changed for the specific use cases of SSC-ICT-Innovatie. It is an example of Retrieval Augmented Generation (RAG) and allows for the use and evaluation of both Open and Closed-Source LLM's. It makes use of Langchain, Chroma and Streamlit among other things to perform document Q&A. Right now it is mostly a clone with some additions (to make running it locally easier for example) and it contains additional project documents (in project_docs) that provide background information about this project, the decisions we made and the research we conducted.
Current repository
The repository is a work in progress of streamlining the larger original Learning Lion Repository and making it more accessible, simple and transparent (for developers and the public) for the current use case. Elements from the original repo are reused. However, we also took the liberty of using newer functions and methods if this would support these higher level goals.
The archive is where the original repo is archived. The jupyter notebook is where we mainly experiment. Later on, we will integrate and abstract the core elements into classes to be reusable and where we will be able to integrate it into (micro-)services.
How to use this repo
The instructions as written assume you have Anaconda and Python installed. If not, download Python (https://www.python.org/downloads/) and follow this installation guide for Anaconda: https://docs.anaconda.com/free/anaconda/install/windows/.