Hey @Nick_D!
I disagree, I think you explained it very well!
I don't think so, I think @Nicholas_Bokulich & I are on the same page as you (more below).
This is what we mean by "per-sample assemblies" --- you have one file per sample containing the assembly sequences. As I mentioned above, in order to dereplicate in QIIME 2 you will need to multiplex (which is a step in the reverse direction) before importing and dereplicating.
For what its worth, I think in the future it will make sense for us to create a new import format in q2-types to support this schema.
Sample workflow
cd path/to/fasta/files
touch merged.fasta
for f in *.fasta; do fn="${$(basename -- "$f")%.*}"; sed "s/^>\(.*\)$/>\\$fn\_\1/" $f >> merged.fasta; done
qiime tools import \
--input-path merged.fasta \
--output-path seqs.qza \
--type 'SampleData[Sequences]'
qiime vsearch dereplicate-sequences \
--i-sequences seqs.qza \
--o-dereplicated-table table.qza \
--o-dereplicated-sequences rep-seqs.qza \
--verbose
qiime feature-table summarize \
--i-table table.qza \
--o-visualization table.qzv