Hello, any feedback on this?

So basically, this is just inputs.exec + parsers.json into one? I don't see the need to add this plugin, while the same can already be achieved by setting some configurations.

As a general comment: the approach I chose was heavily inspired to the nvidia-smi plugin that was already present in Telegraf, so I went straight with that, I admittedly was not aware of the possibility to combine plugins so I can't tell if the combination of plugins you mention is enough to fullfill the case.
A part from the parsing there is indeed some data digestion to produce indirect metrics and tag/fields names, in future iteration this can become more complex. I'll have a look into it.

So basically, this is just inputs.exec + parsers.json into one? I don't see the need to add this plugin, while the same can already be achieved by setting some configurations.

As a general comment: the approach I chose was heavily inspired to the nvidia-smi plugin that was already present in Telegraf, so I went straight with that, I admittedly was not aware of the possibility to combine plugins so I can't tell if the combination of plugins you mention is enough to fullfill the case.
A part from the parsing there is indeed some data digestion to produce indirect metrics and tag/fields names, in future iteration this can become more complex. I'll have a look into it.

I don't mind adding a plugin that just runs a program and parses the json if it's commonly used. Without a plugin for this, each user needs to go through the config process, including learning to use the exec and json plugins and coming up with a good metric format. AMD GPUs are very common so a dedicated plugin could be useful.

If you can quickly make the changes Hipska requested I think we can get this into the 1.20.0 release candidate which comes out Sept 1 (tomorrow).