This is a ChatGPT plugin designed to facilitate browsing of biorxiv.org from https://chat.openai.com.
The plugin enables users to search, download, and extract text of paper PDFs.
Because LLMs have a maximum context length, entire contents of a paper can't be sent directly to ChatGPT all at once.
The plugin trys to overcome this limitation in two ways:
- By letting chatgpt pass the page number to extract text from. You can therefore read the paper gradually to incrementally build context.
- (optional) By providing a langchain based Q&A functionality that converts the entire text corpus to embeddings. You can tell chatgpt to "ask" the plugin specifc questions about the paper, and it attempt to return a response based on matching terminology in a ephemeral vector store.
An OPENAI_API_KEY key is needed for the Q&A functionality. Langchain makes calls to OpenAI so a cost is incurred for each question.
To install the required packages for this plugin, run the following command:
pip install -r requirements.txtTo start the plugin, enter the following command:
python main.pyOnce the local server is running:
Navigate to https://chat.openai.com. In the Model drop down, select "Plugins" (note, if you don't see it there, you don't have access yet). Select "Plugin store" Select "Develop your own plugin" Enter in localhost:5003 since this is the URL the server is running on locally, then select "Find manifest file". The plugin should now be installed and enabled!