Yes, documentation for Flux.1 is definitely a problem, as official sources are often incomplete or outdated. There is an API for working with LoRA adapters that lets you programmatically load and switch between different models at runtime. The main challenge is that different implementations may use varying syntax and sets of available parameters for controlling LoRA adapters. I recommend checking out the API for LoRA models in Flux.1 AI here:
https://yesai.su/en/docs/flux_lora . It includes up-to-date examples and descriptions of the parameters for integrating custom models into applications.