Lucid - An AI Powered Research Assistant

Lucid is built using Streamlit, and powered by Meta's BART (Bi-directional and Auto-Regressive Transformers) model. The BART-Large-CNN model used for this project is especially fine-tuned for summarization tasks.

Usage

Enter your topic of interest in the "Search Query" field on the sidebar and your Semantic Scholar API key below it. If you want to try out the app first, you can use our default API key for 2 free searches before getting your own API. Just select "Use a free search" option and click on the Search button.

Thats it! You would now see a summary of multiple research papers fetched from Semantic Scholar. You can also click on "Read more" to visit the original webpage of that paper.

Features

• Uses MongoDB for storing all the user details
• Offers 2 free searches to users
• Utilises BART Model for summarization tasks
• Stores search history
• Allows users to bookmark research papers

Run the app

Clone the repository on your PC.

git clone https://github.com/ashmit0920/Lucid.git

Install the required packages.

pip install -r requirements.txt

Inside the .streamlit folder, create a file named secrets.toml and add your Semantic Scholar API Key and MongoDB URI.

SEMANTIC_API_KEY = "your-api-key"
MONGO_URI = "your-mongodb-connection-string"

Run the app.

streamlit run app.py

Run using Docker

After creating the secrets.toml file, build the docker image.

docker build -t Lucid .

Run the docker container.

docker run -p 8501:8501 Lucid

Future Updates

• Restructured code using st.Page and st.navigate methods
• Caching using Redis for optimization
• Search history based recommendations
• User profiles