This is a project that uses Google Gemini Pro in tandem with the Langchain framework to read PDFs, extract the text of those PDFs, and then answer queries relative to the PDF. The app was developed using streamlit.
It is recommended that a virual environment is used and the path to the directory containing these files is added to the path in the virtual environment.
These libraries and their dependencies can be installed using the "requirements.txt" file using pip package manager:
pip install requirements.txt- An API key is needed to work with Google's gemini locally. Create a ".env" file with
GEM_API_KEY = "API_KEY"where API_KEY is replaced with the actual API key generated. Be sure to maintain the double quotes.
This pertains to how the text is stored in a way that is speed efficient when looking up a query. Text is stored as an embedding, which a vector of values representing the word or token. In this embedding space, words associated more closely together have more similar embeddings.
A query will be converted into an embedding and then used to query the text to find the most relavant text based on similarity. The resulting text is then fed to our LLM to provide the context that is used to respond to the query.
The method I used was FISS. One can also use other storage methods found here.
To run locally, simply write in terminal (with virtual environment active, if applicable):
streamlit run app.py
From there, connect to the localhost address that the application is running on and it's available for your use!