WebChat

WebChat is a chat bot that extracts and cleans the content of a website, allowing you to interact with it through a conversational interface.

Features

• Ensemble retriever: combines keyword and semantic search to retrieve the source documents.
• Maximum Marginal Relevance: re-ranks source documents maximizing the diversity and reducing redundancy.
• Reorder: places the most relevant documents earlier in the context to avoid the lost-in-the-middle problem.

Installation

1. Set the following environment variables in an .env file in project directory:

OPENAI_API_KEY
WEBSITE_NAME
WEBSITE_URL
WEBSITE_DESCRIPTION

2. Create a virtual environment and install dependencies:

pip install -r requirements.txt

3. Ingest the website content:

python src/ingest

4. Run the streamlit web app:

streamlit run src/app.py

Screenshot and demo video

References

1. LangChain Documentation: https://docs.langchain.com/

2. Streamlit Documentation: https://docs.streamlit.io/

3. Platzi LangChain Course for Document Management and Retrieval: https://platzi.com/cursos/langchain-documents/