Tough one to try and tease apart in regard to frequency range as well as spatial position overlap. That leaves dynamics..
Rather than targeting either instrument individually, it may help to reduce dynamics to increase the perceptual level of the violin relative to the piano. Classic parallel compression will bring up the details without affecting the dynamic peaks, and will tend to sound less obviously compressor-squashed, but will also bring up the noise floor and ambient/audience noise details. Multi-band compression might work. I've had good results with Voxengo's Soniformer plugin that is a somewhat unique take on which incorporates 64 bands using a unique user interface to manage it by drawing curves across all bands.
https://www.voxengo.com/product/soniformer/There are tools emerging that separate sounds based on their attack envelope, separating transients from sustained sound. A link to one of them was posted at TS within the last couple weeks, which is the only reason I mention it, but I can't remember offhand what it's called. Could potentially be applicable if one instrument is playing staccato and the other legato, but it seems a reach to me.
I suspect parallel comp, mulitband comp, or some combination of the two are likely to be your best bet.