Skip to content

Setup geniie-lab

Method 1: Google Colab (Browser) Only

Use setup notebook

  • Download setup_geniielab.ipynb to your Google Drive
  • Open the notebook using Google Colab
  • Follow the instruction

Method 2: VSCode + Google Colab

Check List

  • VSCode has been installed
  • Google Colab Extension has been installed
  • VSCode has connected to Colab Server
  • VScode has mounted Google Drive and you can see a list of files
  • You know how to start a Colab terminal in VSCode

💡 Not yet? Don't worry. Follow the instruction in How to prepare your PC for the hands-on session

Colab Terminal in VSCode

The following instructions should be executed from the terminal in VSCode which is connected to Colab Server.

You should see a prompt like this in the terminal.

/content#

1⃣ Clone geniie-lab dev repository to your Google Drive

cd drive/MyDrive # or other location
git clone https://github.com/geniie-lab/geniie-lab.git -b dev
  • This should create a folder called geniie-lab in your Google Drive.
  • Click Colab icon in Activitity Bar
  • Refresh the file list and find the folder
Colab CPU (or GPU T4)
  ├ drive
      ├ MyDrive
          ├ geniie-lab

2⃣ Setup geniie-lab

💡 You need to do this everytime you make a new connection to the server.

cd geniie-lab
python -m pip install -r requirements_colab.txt
# Collecting ...
python -m pip install -e .
# Obtaining file:///content/drive/MyDrive/geniie-lab
# ...

3⃣ Add .env file

  • Right-click the geniie-lab folder in the file list âž¡ New File...
  • Enter .env as the file name

  • Copy and paste a KEY sent from the organiser

    OPENROUTER_API_KEY="..."

  • Save the file

4⃣ Change the host for OpenSearch

  • Find scripts/run_session_experiment_chiir2026.py

    Colab CPU (or GPU T4)
├ drive
    ├ MyDrive
        ├ geniie-lab
            ├ scripts
                ├ run_session_experiment_chiir2026.py

  • Open the file and change the localhost to an IP Address shared by the organiser

    • Around Line 58
    tools=[
    ToolDescription(
        name="opensearch",
        ranking_model="bm25",
        index_name="scidocs_bm25",
        host="localhost",             # <--- CHANGE HERE
        port=9200,
        use_ssl=False,
        description="It allows you to perform searches using keywords only and employs the BM25 ranking model to order results.",
    )
],
We're good to go!