This is the project for Berkeley CS294/194-196 Large Language Model Agents | CS 194/294-196 Large Language Model Agents

Team Member: Rachel Lin (3035629830) Mutian Hong (3040824725) Chuyan Zhou (3040814117) Jan Chen (3040762169) Lucy Struefing (3040858941)

tested on Ubuntu 22.02. follow the following procedure to build up. The complete architechure is built under mutian-develop-v8 branch. The rag function in the main branch does not works well. So if you want the same architechure in the report, please switch to branch mutian-develop-v8.

Setup

1 clone the code

git clone https://github.com/itsuitsuki/pdf2lec_llm

set up an .env file under root(pdf2lec_llm/) with the OpenAI API Key:

export OPENAI_API_KEY="your API key"

Backend

1. install miniconda (skip if already installed)

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
chmod +x Miniconda3-latest-Linux-x86_64.sh
./Miniconda3-latest-Linux-x86_64.sh
source ~/.bashrc

2. install docker (skip if already installed)

sudo apt-get update
sudo apt-get install -y \
    apt-transport-https \
    ca-certificates \
    curl \
    gnupg \
    lsb-release
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg

echo \
  "deb [arch=amd64 signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu \
  $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
sudo apt-get update
sudo apt-get install -y docker-ce docker-ce-cli containerd.io
sudo usermod -aG docker $USER
sudo curl -L "https://github.com/docker/compose/releases/download/v2.20.0/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose
sudo chmod +x /usr/local/bin/docker-compose
newgrp docker

3. setup opencv (skip if already installed)

sudo apt-get install -y \
    libgl1 \
    libgl1-mesa-glx \
    libglib2.0-0 \
    libsm6 \
    libxext6 \
    libxrender-dev \
    libglib2.0-0 \
    libgtk2.0-dev \
    pkg-config

use tab to select ok

5. setup conda env

conda create -n pdf2lec python=3.10
conda activate pdf2lec
pip install -r requirements.txt
sudo apt install ffmpeg

6. create data folder

cd backend
mkdir -p data metadata

7. start the backend

bash run.sh

Frontend

cd ..
cd frontend

Install Nodejs (skip if already installed)

sudo apt update
curl -fsSL https://deb.nodesource.com/setup_18.x | sudo -E bash -
sudo apt-get install -y nodejs

Start Frontend

npm install
npm run dev

Create Login and use product

Run the application on http://localhost:5173/ should lead you to a login window. click on sign-up and create a user with a password, then log in. Upload your lecture PDF and click on the uploaded PDF. Choose your generation granularity and add textbook (optional). If you uploaded a lecture PDF, it should start loading (~20-30s per slide).