This repository is part of the competition HackMty, where the goal is to create a project to make an efficient and quick generstive AI result, being a search engine.

This project is done to resolve the issue of searching for an scientific paper, articule or report into a big pile with large amounts of documents. Where we only want to obtain information in a certain report that we are looking for. To save up time and energy, this projects has as input one or several article files and a prompt for what is being looked for. Giving as an output the required file and section where the request is obtained from.

The script works with the following versions:

• python 3.11.5

To install the necessary libraries, run the following code in a Python executer

pip install -r requirements.txt

To view the version of your libraries, run the following:

pip show python-dotenv

This project uses Pinecone.io as a Long-Term Memory method for AI. To obtain the necessary information, do the following: Create a Pinecone.io account if necessary and create an Index and API Key, these will be used in the env variables further on.

The .env file needs to have the following environmental variables for the script to work properly:

• OPENAI_API_KEY: API Key you use for OpenAI access
• OPENAI_API_BASE: Base page for the API
• OPENAI_CHAT_MODEL_NAME: GPT model instance being used
• OPENAI_EMBEDDINGS_MODEL_NAME: Model name for OpenAI embeddings
• PINECONE_API_KEY: Generated Pinecone API Key
• PINECONE_ENVIRONMENT: Provided Environment for Pinecone, generated with Index
• PINECONE_INDEX_NAME: Generated Index name for Pinecone