Muffin Mate - Internal HR Chatbot

Youtube Link: https://youtu.be/JgZCWn4texM

Team Members

• Aden Osman
• Copper Wilson
• King Sambonge
• Nathaniel Davis

Project Overview

Muffin Mate is a secure, in-house HR chatbot designed for Muffin Maniacs employees. It provides 24/7 access to company policies, benefits information, and HR guidance, helping reduce HR workload and prevent data leaks. The system leverages a Retrieval-Augmented Generation (RAG) architecture powered by Llama 3 and hosted using Ollama, ensuring that all sensitive information remains within the company’s network.

Key Features

• Natural language processing for HR policy queries
• Fast retrieval of relevant HR documents (PDFs, SharePoint)

Technologies Used

• Ollama
• Python
• Llama3

Setup Instructions for Running the Muffin HR Prototype

1. Download Ollama

Download Ollama from the official site:

https://ollama.com/

2. Install Required Python Packages

Run the following in PowerShell:

pip install llama-index llama-index-llms-ollama llama-index-embeddings-huggingface sentence-transformers

3. Move Project Files

Copy Modelfile and rag_demo.py into your .ollama\models folder. Typically found here:

C:\Users\<INSERTUSERHERE>\.ollama\models

make sure is the user you are using

4. Create the Documents Folder

Inside the .ollama\models directory, create a folder named documents and add the combined_documents.txt

5. Update rag_demo.py

Open rag_demo.py and update line 8 to the appropriate directory from step 4

EX: 

From -
combined_file = r"C:\Users\<INSERTUSERHERE>\.ollama\models\documents\combined_documents.txt"


To - 
combined_file = r"C:\Users\coppe\.ollama\models\documents\combined_documents.txt"

6. Navigate to the Models Folder

Back in Powershell navidate to the directory from step 3 by typing

cd C:\Users\<INSERTUSERHERE>\.ollama\models

Do not forget to change the to the appropriate user

7. Run the Python Script

Still in Powershell, run the RAG demo by typing the following

python rag_demo.py

If the chatbot responds, great! You can keep ask your questions in the command prompt or continue to create a custom model within Ollama to ask the question.

Before continuing, close the model by typing exit or Ctrl+C

8. Build the Custom Model

In Powershell run the following:

ollama create muffin-hr -f Modelfile

Verfiy the model exists by typing:

ollama list

You should see muffin-hr in the list. If you do you should be able to open the Ollama Desktop App and choose it from the dropdown on the bottom right of the app. If not, continue on to step 9

9. Configure the Ollama App

In the Ollama desktop application:

1. Go to Settings
2. Set the model location to:

C:\Users\<INSERTUSERHERE>\.ollama\models\manifests\registry.ollama.ai\library\muffin-hr

3. Go back to the main screen and open the model dropdown in the bottom right corner
4. Select mufffin-hr