Add configurable AI agents

The idea is to add AI agents as a concept. The user may create multiple agents with distinct configurations like entities to "read" triggers to launch the agent, ...

Two examples jump into my mind: 
1. when a new support case is created or a new comment is added by the requester an agent may be launched which is then searching through the requests, bugs, features, user stories, ... of the past x months/years as well as custom website URLs (our product documentation) and the internet to try and find solutions to the case based on other cases solutions. It would also allow us to provide it with a prompt which may contain additional information about our products or something. 
With the result of the research the agent will post a comment (including the sources!) to the request visible only to our staff, not the requester.

2. when a new support case is created the agent will go through the case and detect which product it is about and what kind of request it is: did the user encounter an error, is it a feature request or maybe a request for configuration assistance? depending on the type of request the LLM may suggest the requester to provide a logfile, more detailed error description, screenshots, ... whatever we defined as "required" in the agent's prompt. For assistance requests the agent may look into the documentation for relevant information and share those with the link to the source.