Data collection and intelligent decision are two fundamental factors of success IoT applications or networked applications in general. Many existing IoT applications use cloud services as the main host of collected data and intelligent decision. Examples are SmartThings, IFTTT, Amazon Echo, and Google Home. Using cloud services as the host of intelligent IoT ap- plications have its own merits. The service providers have the full control to the collected data and decision logics. It also reduces the overhead of managing the services, includ- ing revising, updating, and revoking services. Unfortunately, the aforementioned cloud-based service model limits users’ capability on configuring and controlling the services. Revolv market their Smart Home hubs, which lets home users con- trol various appliances and home automation systems with a central application. The company was acquired by Nest in October 2014 and a subsidiary of Alphabet. In Feb. 2016, Revolv and Nest announced that all Revolv services will be terminated, starting at May 15th. Although the owners of Revolv hubs may receive refunds for their hardware devices, all the smart home services have to be migrated, if possible, or will stop functioning. The tradeoff between data privacy and intelligent services is an important issue on developing IoT services. The aforemen- tioned example use cloud-based service model and limits the users on controlling the services. In this paper, we developed a heterogeneous service model, which combines public and pri- vate intelligence to develop IoT applications. It allow the IoT applications to access the cloud-based public intelligence and site-based private intelligence in IoT applications. For the sake of demonstration, nodeRED will be used to compose the applications to access cloud-based public intelligent services and WuKong will be used to compose the applications customized for each deployment.