The general meaning of an agent is "something that is capable of producing a particular effect".
In the context of Embedia, an agent has the following aspects:
- It uses a large language model as the decision maker
- It is designed to accomplish a specific task
- If given the access, it can modify things in its environment (using
- It runs in a loop, taking decisions (using the LLM) based on its current state and acts on them to accomplish its task
Embedia's agents are designed in a way that they can be controlled by other
agents. Hence, every agent will be a subclass of the
Tool class. (Learn more
about: Tool) In that sense, the parent agent
will use the child agents as tools to accomplish its task.
The prebuilt Agents that are shipped with the library are: