Exploring the principal libraries
After discussing the various components and frameworks on a conceptual and theoretical level, in this section, we will discuss some of the major libraries that allow these concepts to be put into practice. These libraries make it possible to connect an LLM to the various additional modules. LLMs remain central but are thus connected to perception modules and execution tools. In the next chapters, we will elaborate on some aspects and see different practical examples.
In general, the structure of an LLM-based application has several components:
- The interface: this connects the user to the system.
- The brain: an LLM that can also be connected to additional memory. An LLM has its own parametric memory (obtained during training), but we can add external memory (such as a vector database or knowledge graph).
- Perception modules: these allow the ingestion and transformation of user data.
- Tools: modules that extend the abilities of the...
