Note that this won't give exactly correct results when the sensor is being accelerated by an external force (when you are pushing or rotating it), because there are other unknown forces besides gravity acting on the sensor. The mpu6050 sensor platform allows you to use your MPU6050 Accelerometer/Gyroscope (datasheet, SparkFun) sensors with ESPHome. Using this knowledge, we can find how much the sensor has pitched or rolled, by calculating what component of the gravity vector has moved from the negative Z axis, to other axes. However, how should the position of an object fitted with accelerometer and gyroscope be determined using the data from the. When the sensor is sitting at rest, it will sense the force due to gravity in the negative Z direction as 9.8 m/s^2. The double integration of acceleration gives the position of an object. Pitch and roll Euler angles in the global frame can also be calculated with the accelerometer. This helps us to measure acceleration, velocity, orientation, displacement and many other motion related parameter of a system or object. Once you have the angular rates in the global frame, you can then integrate and accumulate the angles as mentioned in other answers. The MPU6050 is a Micro Electro-Mechanical Systems (MEMS) which consists of a 3-axis Accelerometer and 3-axis Gyroscope inside it. Sections 9.1, 9.2, and 9.3 in this OCW pdf lays out how to do this pretty well:

So the first task (before integrating), is to convert the angular rates from the sensor-body frame to angular rates in the global frame. In other words, the X-Y-Z coordinate axes of the sensor are spinning with the sensor, whereas the global X-Y-Z axes remain fixed. So if the sensor is rotating with some rate, the data it's outputting will be in the frame of the rotating sensor, NOT with respect to the fixed global frame. First thing to note is that the gyroscope is reporting angular rates with respect to the sensor.

Accelerometer gyroscope