Skip to content

ChatKNML is a Python-based chatbot powered by Large Language Models (LLMs) with an integrated Discord interface designed to respond to queries related to the Machine Learning Science Club and Rzeszów university of technology.

License

Notifications You must be signed in to change notification settings

knmlprz/ChatKNML

Repository files navigation

ChatKNML

KNML

Python Docker Poetry Django Discord.py PostgreSQL Git MIT License

Getting started

First setup

  1. Clone the repository:

    git clone [email protected]:knmlprz/ChatKNML.git
  2. Navigate to the project directory:

    cd ChatKNML
  3. Create a new branch for your development work:

    git checkout -b {your_branch_name}
  4. Make sure you are working on the correct branch:

     git status

Starting app development

  1. Copy the .env.example file:

    cp .env.example .env

    Modify the environment variables to suit your requirements.

  2. Launching services using the "dev" profile:

    docker compose --profile dev up

Starting app production

Starting app

docker compose --profile prod up

Starting llm and embedding

  1. Download model (must have for service llm-embedding to work!!!)

    Download model (size of file 3.6GB ):

    curl -o ./llm/models/llama-2-7b.Q3_K_L.gguf -L https://huggingface.co/TheBloke/Llama-2-7B-GGUF/resolve/main/llama-2-7b.Q3_K_L.gguf

    or

    wget -P ./llm/models/llama-2-7b.Q3_K_L.gguf https://huggingface.co/TheBloke/Llama-2-7B-GGUF/resolve/main/llama-2-7b.Q3_K_L.gguf
  2. Launching llm and embedding

    2.1. Running on cpu

    docker compose --profile cpu up

    2.2. Running on gpu

    docker compose --profile gpu up

LLM and embedding api swagger

Swegger with EP for completions(llm + embedding) and only embedding is here

About

ChatKNML is a Python-based chatbot powered by Large Language Models (LLMs) with an integrated Discord interface designed to respond to queries related to the Machine Learning Science Club and Rzeszów university of technology.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published