dfrobot INA219 Gravity: I2C Digital Wattmeter

AI context

Dense context for pasting into a model prompt — addresses, pinout, voltages, tolerances and the vendor's reference circuit.

PART: DFRobot Gravity: I2C Digital Wattmeter, SKU SEN0291, 30.0 x 22.0 mm, 4 g. Built on a Texas Instruments INA219 zero-drift current and power monitor with a 0.01 ohm (10 milliohm) low-temperature-drift alloy shunt, on a Gravity carrier with a 3-pin screw terminal for the load path.
I2C ADDRESSES: four options, 0x40, 0x41, 0x44, 0x45, selected with the on-board DIP switch that the pinout table calls position 5 'ADDR - I2C address DIP switch'. Those are the INA219's A0/A1 combinations (both low, A0 high, A1 high, both high), so four meters can share one bus; any other address needs rework.
PINOUT: 1 VCC (3.3-5.5 V), 2 GND, 3 SCL, 4 SDA, 5 ADDR (I2C address DIP switch), 6 3P TERMINAL - the KF396 3-way screw terminal that carries the load path through the shunt.
SUPPLY AND MEASUREMENT: VCC 3.3-5.5 V, quiescent current 0.7 mA. Bus voltage 0-26 V measured at IN+ or IN- relative to GND; current 0 to +/-8 A bidirectional; power 0-206 W. Resolution 4 mV, 1 mA and 20 mW in hardware (4 mW when computed in software). Voltage error < +/-0.2% typical; current error < +/-0.2% typical only after a manual calibration, which DFRobot states is required before use.
LOGIC LEVEL: DFRobot's spec table says 'Gravity I2C (logic level: 0-VCC)', but the module's own schematic regulates VCC down to 3.3 V with an ME6206A33M3G and feeds the INA219's VS from that rail, so the bus runs at 3.3 V whatever the host applies. Both statements are vendor-sourced and they disagree; the schematic is the safer read.
REFERENCE CIRCUIT: U1 INA219, U2 ME6206A33M3G 3.3 V LDO with 1 uF on its input (C1) and 1 uF on its output (C2), R1 0.01 ohm shunt, and four 10 k resistors (R2, R3, R4, R5) on the I2C and address nets. Two 4-pin I2C connectors are wired in parallel - a Gravity socket (P2) and a plain 4-pin header (P3) - and P1 is the 3-way KF396 load terminal.
TRAPS: (1) The shunt is high-side: IN+ and IN- sit in the positive lead of the load and are not interchangeable, and current is reported as bidirectional so a reversed load reads negative. (2) 4 mV and 1 mA are resolutions, not accuracies - the +/-0.2% current figure only holds after calibration. (3) The 26 V and 8 A limits are the chip's, not the screw terminal's; the terminal and the board's own copper are what you should judge first.