DocsChatBot2.mp4
A chatbot built using LangChain and Gemini that answers user queries from documents and includes a conversational form for collecting user information. Features date extraction and validation for user inputs.
It is a implementation of document retrival type chatbot that can answer based on information available in a python file. Currently I am using Attention is all you need paper. This file is kept in sample_docs folder as attention_is_all_you_need.pdf.
You can Change the content of file or use multiple files with simple tweek in the code.
Now Booking ststem is also implemented. Booking session is initiated by simply asking to make a booking in Natural language.
Booking system can extract information from the natural conversation between user and ChatBot. This System can efficiently collect information in conversational Form.
Clone the repository in a folder.
git clone https://github.com/AnjaanKhadka/DocsChatBot.git
Get in the direcory
cd DocsChatBot
All dependencies are kept in requirements.txt file.
pip install -r requirements.txt
Beside requirements, you need to have Gemini API key. You can create a free Gemini API key for testing purpose.
After getting the key, Copy your API key and paste it in your API_KEY.txt file in the primary direcory.
echo Your_API_KEY > API_KEY.txt
I am using Streamlit to create interactive chatbot. Run the main.py with streamlit.
streamlit run main.py