Thank you, somehow I didn't find that page before or didn't understand how useful it is to my problem.
Nevertheless, in the meantime, I did a blastp pairwise alignment and I leave the steps here as it might be useful to someone. It is a half baked solution as it seems to duplicate the workload (the same sequence pairs are calculated twice) and it automatically chooses only sequences that produce significant alignments (not all sequences you provide)
1 - prepare a database of all desired sequences
makeblastdb -in All_seqs.fasta -dbtype prot -out my_prot_blast_db
2 - pairwise alignment
blastp -db my_prot_blast_db -query All_seqs.fasta -out all-vs-all.tsv -num_threads 8
I will now try Biopython’s command line wrapper for EMBOSS