As an avid reader of research articles in fields ranging from electronic structure theory all the way to the machine learning methods and the application of everything that encompasses in between for energy storage materials, I have a lot of reading to do. I am always downloading research articles and taking screenshots of papers in my phone while checking email or browsing social media. (Yes, LinkedIN and Twitter are a great place to be updated). Eventually, reading lists keeps increasing and tabs just never end. 

So, using my awesome Mendeley Desktop Library which I have been maintaining for 5 years now, I have a special folder in it called 'need sorting' where unread and unorganized papers are stored. I made a Gemini based LLM summarizing tool which reads all those papers for me using Mendeley API interface and stores a 200 word summary for me. I tested responses from Claude, OpenAI as well but Gemini Flash model turned out best. Then I made a recommender system, where I prompt one out of 24 available models using API asking for a topic I want to read and it recommends me a n number of papers based on my prompt. It is also customizable, in a sense, you can tell the model the response wasnt good, or want to try a different model for recommendation, or give me more information why did you recommend this paper, or try search again with more information. Finally, if you like a paper, you can delete that paper from the database and continue adding more as your list of unread papers increase. 

Go check it out on my github or gitlab and let me know what you think.