[dalke]

I looked at the PDB parser.

Will you be adding support for using duplicate CONECT records to store bond type information? That's a RasMol extension that PyMol supports. You'll also need to support the pdb_conect_nodup option in the writer.

I see you interpret atom/hetatm serial numbers as an integer. Will you be using the base-36 or hybrid-36 variants (see https://cci.lbl.gov/hybrid_36/) which is a common way to handle more than 100,000 atoms?

Again, these are corner cases which come with experience. I've no expectation that a new program would handle them. I want you to know about them since they will be issues if you expect long-term uptake.