[Design] [Gen AI] Overall refactoring (settings location, ease instanciation on buildin stories in python..)

[Benvii]

Main idea is to think of an overall refactoring of the Gen AI related settings focus on several axis :

• Do not store everything on the Bot Definition
• Provide those settings for build-in stories a add a way from tock-py to access such settings to ease the use of langchain custom chains implementation inside buildin stories (easily access or build an LLM, embedding ...)

As each bot can have specific needs that would take advantage of Generative AI and especially langchain we would like to keep generative AI settings at the Studio level, but not only for RAG and and let developers easily take advantage of those settings and maybe factory to build langchain chain using tock-py.

---

          > Thanks for adding all the check document compressor settings to the orchestrator 👍️, juste a small comment about an exception raising that could be improved.

We begin to have a lot of Gen AI related configuration directly on the Bot Definition, I'm not sure but in a futur refactory I might be a good option to group them especially if we want to expose the to python scripted stories (tock-py) and also kotlin scripted stories. It would be great to have access to some BotDefinition configurations inside buildin sotries (stories behind the Bot API). Maybe we should create an issue about that. My idea is in a near futur to expose Gen AI settings to build in stories (especially python, tock-py ones) so that python developpers could easily build there own langchain chains using the bot settings and the gen-ai-orchestrator factories provided in a separated python package.

We also have a lot of gen ai related configuration endpoint directly on the bot admin verticle we should in the futur group them in a different verticle.

Yes, we need to include it in the analysis when we refactor the orchestrator code. We need to put this in a separate ticket.

Originally posted by @assouktim in #1788 (comment)