So, allotetraploid means that "heterozygosity" will be very high, right? If so, you should have a high enough sequencing coverage.
Before you start assembling a genome, do a basic QC on your reads with GenomeScope2.
For genome assembly you can try Hifiasm (use the --ont parameter for Nanopore data). I have only used Hifiasm with PacBio reads, so I'm not sure if it works equally well with ONT.
For QC of the assembly make sure you run BUSCO (for gene completeness), Merqury.FK (QV, k-mer completeness), and BlobTools (for spotting any contaminants).
Before gene annotation, you should mask repetitive regions in the genome; use RepeatModeler to find repeats appearing only in your genome (i.e. species-specific repeats). Then, use RepeatMasker to mask the species-specific ones from RepeatModeler and also the conserved ones from the Dfam database (see the RepeatMasker page for details on how to install Dfam).
For gene annotation I've used BRAKER. Installation could be a pain, but I see that they also have containers, so it probably is easier. Make sure you have RNAseq data from the same individual (or species, at least) if you want to do a decent job. And don't forget to run BUSCO on the predicted gene set. And then compare the two BUSCO scores (from the genome assembly and from the predicted gene set) to see if you're missing any genes. If you're missing genes, then the RNAseq or protein evidence you used for BRAKER weren't enough. In such a case you should repeat gene annotation using e.g. additional evidence (more RNAseq data, or more protein evidence).
Do you know if your nanopore data was basecalled using fast, high or super accuracy models? Generally the higher the accuracy of basecalling, the better all downstream analysis is going to be.
I suggest hiring a bioinformatician or taking a bioinformatics/genome assembly course. Just the assembly of an allotetraploid genome using nanopore reads is not an easy task. I seriously doubt you'll get the proper guidance in a single post.