A Discord-based chatbot designed to answer university-related questions using LlamaIndex and Llama 3.1. Fork of https://github.com/TillHaenisch2/chatbot_poc

1. Features
2. Technologies Used
3. Installation
4. Usage
5. Adding and Managing Sources for RAG

• Answers university-related questions based on faculty guidelines and other documents
• Integrates seamlessly with Discord for easy access
• Utilizes LlamaIndex for efficient document indexing and retrieval
• Powered by Llama 3.1 for natural language processing

• Python
• LlamaIndex
• Llama 3.1
• Discord.py
• Docker

The "prooven" way to install the chatbot is via ansible on a linux system. The alternative way is to install the chatbot manually. Therefore, the playbook can be used as a reference. The installation playbook is located in the infrastructure directory.

1. Install Ollama https://ollama.com/

2. Pull model

   ollama pull llama3.1:8b-instruct-q6_K

3. Install Python requirements

    pip install -r requirements.txt

   there may be missing some packages, just install them with pip

4. Install docker https://www.docker.com/

The chatbot can be configured using environment variables. The following environment variables are available:

Variables with no default value are required.

Start the redis data base by running following command:

docker compose up -d

To start the chatbot, run the following command:

python Bot_main.py

To manage the sources for the bot which can be used by the bot to answer question the RAG solution is choosed. All files to include to the vector store must be placed in the data/[course] folder. For each course one sources.json is needed. This file has the structure shown below:

{
  "sources": [
    {
      "priority": 1,
      "name": "Leitlinien Studium",
      "file": "Leitlinien.pdf",
      "description": "Informationen über die Studienordnung, Prüfungen",
      "web_link": "https://daten.de/Leitlinien.pdf"
    }
  ]
}

https://docs.cloud.llamaindex.ai/llamaparse/features/supported_document_types

.docx .pdf .xlsx .txt .csv

• Error loading "fbgemm.dll" or one of its dependencies:
  • Solution: Install the Visual Studio Installer and the MSVC v143 VS Build Tools
    Read more here