Very nice 3D print design!

For me that looks like the heater is a later addition to an existing system, so there was only one pipe to the shower. The tilted mount of the kitchen sink faucet emphasizes this for me. :-)

But you could easily improve your system to thermostatic control. Add a sensor on the pipe to the shower (directly to the right side of the hot water tap for the kitchen) and let the ESP on the valve servo read it's temperature and control your mixing valve (the bypass of the heater) to the desired temperature. I would much more prefer an automated system than to have to fumble on the phone with wet hands to adjust the temperature. Plus the sensor can be much closer to the heater so you have much less lag in the control loop than in your manual control loop.

If you don't have the ADC input of the ESP-module available then use a digital temperature sensor like a DS1820 or a I2C or SPI sensor. Probably you need some indication of water flow. Perhaps you do not want to open the valve completely to increase the output temperature only because the heater is not running because no water is used. But perhaps this is no issue as the water heats up gradually when it is turned on and the gas flame starts burning. For that indication - if needed - I would either try to sense the noise of the flame (if possible) or sense the heat of the exhaust gas.
 You can also sense the hot water output for temperature increase

some good ideas there . i was thinking of putting a remote sensor on the shower rather than the heater as there is a lag of almost 2-3 minutes before the actual hot water ( after losing some heat due to the insulated gi pipe running  5 meters ) reaches the shower. This way it can send a remote signal to the base to shut the valve if the temperature exceeds a set limit.. The sensor can plugin to the remote so it could do a dual job with or without the sensor.. i didnt know that the esp had a adc ( i am using the nodemcu ) i will check the pinouts again.. if it does then a thermistor could do the trick else i will look for a adc and a thermocouple

This lag - I would have expected seconds not minutes for 5m - is a killer for an efficient and stable control loop. Just: think: A sensor on the shower had senses too hot water and turns the valve ( according to your pictures it controls the bypass valve, so it has to open it to reduce the temperature). But your pipe is already full of hot water, which temperature can by no way influenced any more. So your control loop needs to be incredibly slow, wasting copious amounts of water or it oscillates like crazy.

If you have the sensor as close as possible to the heater (after your control valve) there is only very little water between your way of influence (control valve) and the sensor. The loop can be much faster and stable.

The ADC input of the ESP has an unusual name (TOUT) which in no way lets you think of an ADC. Its pin 2. The full scale voltage is 1V and it's not extremely precise. I think it's some kind of "byproduct" of the RF frontend circuitry. There some voltages have to be sensed for RSSI and internal calibrations of the RF circuitry. We used it to sense the battery voltage in a project for state of charge indication.

I think the water heater itself introduces also some lag after turning on the faucet until it heats up. So it could be necessary to use a second sensor directly at it's output to give the control algorithm information if and how hot water is available.

Great idea to control a tap without "intrusions"!!

Any chance to get the 3D project files...? (I' don't have a 3D printer, but this time may be the occasion to get one...)

sure but they are very site specific so wont be of much help. i can give you the source file which are in rhino if you want.

thanks