sondreum / gen-ai

FYS-3810 24V Individual special curriculum Master's degree - Generativ AI

Generative AI - course project

UiT The Arctic University of Norway
FYS-3810 Generativ AI 24V Individual special curriculum - Master's degree

• ⚠️ This project is a proof of concept, and is not intended for production use.

Table of contents

• Generative AI - course project
  • Table of contents
  • Project description
    • Project activity
  • Installation
    • API keys
      • Optional keys
  • Usage
  • Project structure
  • Disclaimer

Project description

An OSINT (Open Source INTelligence) tool that creates a report about a chosen organization, company or individual.

The tool can extract information with the following targets

• Company
• Organization

More targets may be added in the future. Other targets may work, but are not guaranteed or supported.

The data is then used to generate a report containing information about the target. The report may contain the following information:

• Location/Address
• Contact information
• Social media accounts
• Recent activities

Project activity

Installation

Clone the repository in cli or download the zip file from the github repository.

git clone [email protected]:SondreUM/gen-ai.git

To run the project, you need to have python3 version 3.10 or newer, and python3-pip installed on your system. Then, you can install the project dependencies by running the following command:

pip install -r requirements.txt

Because of the size of some of the dependencies, we recommend using a virtual environment to install the them. Creating a virtual environment should be done before installing dependencies. You can create a virtual environment by running the following commands:

# creates the virtual environment
python3 -m venv venv
# activates the virtual environment
source venv/bin/activate

Deleting the virtual environment folder will uninstall all dependencies from the project.

API keys

The project requires a gpt api key to be provided as plain text in the src/keys/gpt_key.txt file. The key can be obtained from https://beta.openai.com/signup/

Optional keys

• src/keys/1881_key.txt - 1881.no API key

Usage

To run the project, you can run the following command:

python3 src/processing.py -e <insert entity>

---

Project structure

Disclaimer

This project is for educational purposes only. The authors do not take any responsibility for the use of this project.

gen-ai's Issues

Modularity of LLM used

Convert gpt into a general LLM class for easier integration with alternative models

Open up integration with ollama for using local models

enable multiple crawls

Currently only a single crawl is supported

Create filter/middleware for gather layer

Create filtering mechanism that reduced the amount of output from gathering layer to reduce layer.
Parse html into markdown or txt that can be used in a LLM

"gule sider" getter

Retrieve information from gule sider

Deciding LLM models

Find suitable models for the project.

1. Summarization Able to generate a True/correct summary from the provided source data. The model must be able to deduplicate the same data from different sources. The model should provide the output in a suitable format such as plain text or markdown.
2. Generating The model should create a human readable report from the given data.
3. Q&A using the gathered data either from source or summary from previous model, it should be able to answer questions about the target subject.

Create abstract class for api modules

brønnøysund register getter

Create verify module

A module that takes the output from the consumer LLM and cross-references with the raw data. This is to prevent instances of LLM hallucination.
These values should be verified that they exist in the source data.

• mail addresses
• phone numbers
• org-numbers
• etc

Converse about report

add the possibility of asking the LLM about the contents of the report