...would some sort of peak-hold circuit help?
No. For estimation may be, like indicator "load on" "load off" - not measuring device.
There are a lot of confusion I see. You have to set a priority list, what is more important, what is less.
Don't mean to be a mentor, but it isn't an excuse "i don't have a pin, right resistor or good ADC". If there is no resources, space , board etc, don't measure at all, put dummy blinking led or LCD and show random lighting effects.
IMHO, accuracy should be starting point (let say 0.5%), than if other factors - price space availability are not match, you degrade your initial goal (1%? - 3%?) and re-design.
Start from avr465, very little resource demanding and low costs design.
Not enough pins - put port extender.
I saw online that basically just did a voltage divider, capacitor, and offset circuit.
More than enough to make arduino ADC happy. You don't need precise reference DC voltage for AC measurements. In first post I've seen TL072 - ditch it, two resistor o'k.
I hadn't intended to modify the original project I'd found beyond changing to potentially using the ADS1115 (and probably throwing a 5v1 zener on the pins of the ADS1115 to protect it - would that be a good idea with a CT?)
Bad idea. Follow reply #2, put a resistor in series with adc input. Data sheet ads1115 says 10 mA input current at max, so select right value of resistor. Don't use a zener leakage current 'd ruin your objective of accuracy.
Out of curiosity, how accurate can I be using an ADS1115? For the most part, I'm hoping to be accurate to around an amp if possible. I'd planned on doing 100 reads of each leg (alternating between the two) as fast as it would let me do the reads, and averaging each leg out. And then doing that polling about every 30 seconds or so. I don't know how accurate it would be overall,
Poling ones in 30 seconds? So you designing an indicator? To measure , sampling must be continuous. Interrupt driven - exact in timing.
Now sampling rate .
Industry standard (Can't find, probably IEC1036.) prescribe up to 20-th harmonics, 60 x 20 = 1200 and x2 for Nyquist = 2400 minimum sampling rate.
You should understand, that accuracy depends on 3 factors:
1. continuous fast sampling;
2. resolution of the ADC + PGA
3. software - processing data.
Having 16 or even 24-bits ADC w/o accounting other 2 factors - doesn't make any sense.
Locate AN578 application note by AD. It explains why do you need 17-bits resolution , fast ADC.
Slow 16-bits ADC provides less accuracy than built-in 10-bits Arduino's with PGA (AVR465).
Out of curiosity, how accurate can I be using an ADS1115?
If you do sampling exact in timing, 860 SPS by ADC1115 gives up to 6-7 harmonics, so accuracy would depend on AC current content. For resistive load, you can get 0.1% I think, but modern switching power supply as a load, deteriorate this value to 1-3% or even less.
I'd planned on doing 100 reads of each leg (alternating between the two) as fast as it would let me do the reads, and averaging each leg out.
You mean averaging squared samples? This is number 3, software, most important part.