Summary

In this chapter, we approached the end-to-end implementation of a conversational application, leveraging LangChain’s modules and progressively adding layers of complexity. We started with a plain vanilla chatbot with no memory, to then move towards more complex system with the ability to keep trace of past interaction. We’ve also seen how to add non-parametric knowledge to our application with external tools, making it more “agentic” so that it is able to determine which tool to use, depending on user’s query. Finally, we introduced Streamlit as front-end framework to build the webapp for our Globebotter.In next chapter, we will focus on a more specific domain where LLMs are adding values and demonstrating emerging behaviours, that is the one of recommendation systems.