Know Your PDF is an advanced system that leverages the power of Retrieval-Augmented Generation (RAG) to provide intelligent chat capabilities for interacting with PDF documents. It integrates four cutting-edge models: ChatGPT-4o, Gemma-8B, Mistral-8B, and Llama3-8B, each bringing unique strengths to the table.
Features
-
RAG-Based Interaction RAG combines the strength of retrieval mechanisms and generative models to provide accurate and contextually relevant responses. This allows Know Your PDF to:
- Retrieve Information: Extract specific data from your PDFs.
- Answer Questions: Provide detailed answers based on the document's content.
- Summarize Content: Generate summaries for sections or entire documents.
- Generate Insights: Offer insights and explanations derived from the text.
-
Multi-Model Support By supporting four different models, Know Your PDF ensures versatility and robustness in handling various types of queries:
- ChatGPT-4o: Excels in generating conversational and contextually rich responses.
- Gemma-8B: Specialized in nuanced understanding and detailed explanations.
- Mistral-8B: Known for its efficiency and accuracy in information retrieval.
- Llama3-8B: Offers superior performance in processing large and complex texts.
The performance of the models was evaluated on the following metrics:
1. Similarity Score: The cosine similarity between the generated Original Response and LLM Response.
MODEL: ChatGPT-4o SCORE: 0.83
File : Evaluation
- Clone this repository
- Create a folder .streamlit and create a file secrets.toml and add your API KEYS
HUGGINGFACEHUB_API_TOKEN="<KEY>"
OPEN_AI_KEY="<KEY>"- Open app.py, Change the model you like
....
....
# - Mistral0.2-7B
# - Gemma1.1-7B
# - ChatGPT-4o
#USE THIS EXACT NAMES
model_choice = "ChatGPT-4o" #CHANGE ME
st.set_page_config(
layout="wide",
page_title="Know Your PDF",
page_icon=":book:",
initial_sidebar_state="expanded",
)
....
....Just type this following in your terminal
cd know-your-pdf
streamlit run app.py