Summary

In this chapter, we introduced the concept of multimodality and how to achieve it even without multimodal models. We explored three different ways of achieving the objective of a multimodal application: an agentic approach with a pre-built toolkit, an agentic approach with the combination of single tools, and a hard-coded approach with chained models.

We delved into the concrete implementation of three applications with the above methods, examining the pros and cons of each approach. We saw, for example, how an agentic approach gives higher flexibility to the end user at the price of less control of the backend plan of action.

Finally, we implemented a front-end with Streamlit to build a consumable application with the hard-coded approach.

With this chapter, we conclude Part 2 of the book, where we examined hands-on scenarios and built LLMs-powered applications. In the next chapter, we will focus on how to customize your LLMs even more with the process of fine...