Retrieval-Augmented Generation Experiment

An initial exploration of retrieval-augmented generation for a research software repository.

Quickstart

1. Clone the Repository:

   git clone <repository-url>

2. Add the API URL to Your .env File:

   • Add the following lines to your .env file in the root directory:

     API_URL="Your API URL here"
     QUERY="Your query here"
     

3. Navigate to the Project Directory:

   cd <project-directory>

4. Install the Necessary Dependencies:

   pip install -r requirements.txt

   Using Python 3.12.4 is recommended.

5. Create the directories:

• Create a /models, /vectorisations and /data directory in the root directory.

6. Create the Dataset & Vectorizations:

   • Before running the notebook for the RAG experiment, you need to create the dataset and generate text vectorizations for the retrieval part of the RAG.

   • To do this, simply execute the 1_vectorisation.ipynb notebook. The data will be saved to your machine and will be available the next time you open the project.

7. Install Ollama

• Install Ollama and download the model you want to use.
• To install llama3, the model we're using in this notebook, run the following command ollama run llama3
• Ollama has to run in the background for the chat-bots to work.