jan loads/unloads the LLM model into memory for each query

[MRDOCTOROO]

Because it is unloaded every few seconds when not used, it takes a long time to load the model each time, and the cost is higher. Also, the model list cannot be obtained using the API service. When the model is not started, a whitelist is added and it is started with 0.0.0.0. The model cannot be accessed from outside the local cross-domain.