Hi @akknight216, before I address the error you brought up, I think it is worth mentioning that it looks like maybe you are combining multiple sequencing runs prior to denoising --- this isn't recommended (more on why in a bit!).
Our FMT tutorial provides a good example of running an analysis on multiple sequencing runs. Basically, you import the demultiplexed data on a per-run basis (for example, our FMT tutorial has two runs, sounds like you might have more). Then, you can run qiime dada2 denoise-paired on each run, and finally merge the results. Once everything is merged, all of your downstream analyses should proceed as if you only had one sequencing run (as in, you shouldn't need to do anything different).
The DADA2 docs have this to say about why we use that strategy:
Large projects can span multiple sequencing runs, and because different runs can have different error profiles, it is recommended to learn the error rates for each run individually. Typically this means running the Sample Inference script once for each run or lane, and then merging those runs together into a full-study sequence table. If your study is contained on one run, that part of this script can be ignored.
So, in the short-term, it seems like you can keep moving forward with your analysis by adopting this per-sequencing-run strategy.
As far as the ValueError: paired-end-demux.qza is not a QIIME archive error, would you be willing to share a minimum data set that reproduces the problem? If so, I can follow up with you in a direct message about how we can transfer the data or arrange access to your server.
Thanks!