Feature request: Support dual-model routing (Fast/Lightweight vs. Deep/Reasoning

[666HELPER666]

Description: Is it possible to implement a hybrid multi-model architecture that automatically (or semi-automatically) routes requests (prompts) based on the required level of complexity? The core idea is to combine two types of models to optimize speed, cost, and response quality: a lightweight, fast model (e.g., Qwen 3.5 2b) for simple interactions, user greetings, or basic tasks; and a "smart" model with advanced reasoning capabilities (e.g., Qwen 9b) for complex requests requiring deep analysis, coding, or multi-step logic.

[0xSelenicDove]

AIRI is primarily designed as a companion, not a deep-analysis or coding assistant, so I’m not sure this kind of dual-model routing is necessary for its core use case.