A streamlit app for keeping up with LLM related research.
Up-to-date coverage: https://gist.github.com/masta-g3/8f7227397b1053b42e727bbd6abf1d2e
Everything is automated and stored in a DB.
- New paper title gets added to https://gist.github.com/masta-g3/1dd189493c1890df6e04aaea6d049643.
- Paper meta-data and content is fetched via the
arxiv
library and langchain'sArxivLoader
. - GPT runs reading and summarization process over paper content, generating template of output review.
- BERTopic model is run over full paper set to generate topic groups and labels.
- SDXL pipeline runs to generate paper thumbnail.
- Streamlit app is updated and deployed with new content.
- Custom ComfyUI (https://github.com/comfyanonymous/ComfyUI).
- All dependencies under
dev_requirements.txt
. - Custom
ArxivLoader
(to avoid rate limiting).
The following environment variables are required to run the app (API keys + DB config):
OPENAI_API_KEY
COHERE_API_KEY
HUGGINGFACE_API_KEY
DB_NAME
DB_HOST
DB_PORT
DB_USER
DB_PASS
A populated database is also required to run the app; instructions for setting it up coming soon.