• English
• 日本語

GenSurv (Generative Survey) is a volunteer initiative aimed at exploring and establishing methods to streamline literature reviews using generative AI. Currently, we are engaged in the following specific endeavors:

• Developing an agent that engages in dialogue to delve deeper into the information users seek, retrieves appropriate literature, and provides summarized responses
• Visualizing areas not approached by prior research by vectorizing literature information
• Developing a system that stores literature information in a vector database and uses it as a basis to answer user questions

For more details about the project, past activities, and how to participate, please check here.

• Python 3.11.4

Environment variables

cp .env.sample .env  # Replace with your own API keys, etc.

Python environment

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Executing the script

python src/main.py --title "Laboratory automation" \
  --retrieve_papers --max_papers 5 \
  --generate_headings \
  --output_path data

To run the evaluation function（evaluate_headings.py） from the src directory, use the following command

python -m gensurv.scripts.evaluate_headings \
  --input_data_path ../data/test/manual_by_ono/headings_input_data.json \
  --eval_data_path ../data/test/manual_by_ono/headings_evaluation_data.json

Launching the application (locally)

gradio src/app.py

Launching the application (creating a shared link)

gradio src/app.py --share

graph TD
    A[Start] --> B[Parse Arguments]
    B --> C{Retrieve Papers?}
    C -->|Yes| D[Retrieve Papers]
    C -->|No| E[Load Papers]
    D --> F{Generate Headings?}
    E --> F
    F -->|Yes| G[Generate Headings]
    F -->|No| H[Load Headings]
    G --> I[Classify Papers]
    H --> I
    I --> J[Generate Overview]
    J --> K{Generate Draft?}
    K -->|Yes| L[Generate Draft]
    K -->|No| M[End]
    L --> M