Inspiration
The idea was to develop a chatbot that helped customers and engineers working in bunq having a better time searching for information in the bunq API documentation.
What it does
It is a RAG based chatbot that, given an input from the user, retrieves the most coherent parts of the API documentation, and generates a structured answer.
How we built it
Backend: LangChain, Llama 3.1 8b model for LLM, BeautifulSoup, Selenium Frontend: Streamlit and Streamlit Community
Challenges we ran into
Document scraping from web, setting up the rag pipeline and prompt engineering.
Accomplishments that we're proud of
Cool and usable frontend, available both for desktop and smartphont, and fully working chatbot.
What we learned
Markdown based RAG technology, Streamlit frontend development and deployment, Web scraping.
What's next for bunqbuddy
Integration with Confluence and internal bunq code documentation.
Built With
- beautiful-soup
- github
- langchain
- python
- selenium
- streamlit
- vscode
Log in or sign up for Devpost to join the conversation.