Configurable prompts and input preprocessing for AI assistant features

Such general context information are part of future ideas.

About the context window, when this length is also the value which is configured for the Ollama setup (or any other setup), then the window should be fine, it needs to be a very very very long ticket to reach this.

I can maybe check during the week, if this links are some general problems.
Maybe the LLM size can be a problem related to the JSON return structure, difficult to say.