I think I understand build_spec_generator parts, but would need to understand the database bit in the larger context in more detail. My understanding is like this: We have a BaseBuildSpecDict, which is a typed dictionary storing inferred build data. We define BaseBuildSpec as an abstract base class which requires the resolve_fields method. MavenBuildSpec and PyPIBuildSpec implement BaseBuildSpec and will have their custom logic which populates BaseBuildSpecDict. From reading through the code, the most important functions seem to be src/macaron/build_spec_generator/build_spec_generator.py#gen_build_spec_for_purl , inside where we get a "session" from which we can query build data that macaron has inferred (this is the bit I should understand better). This session is passed to src/macaron/build_spec_generator/core.py#gen_generic_build_spec alongside the purl and patches, which then populates the base BaseBuildSpecDict, before looking up the ecosystem and calling resolve_fields. Back in gen_build_spec_for_purl, format for the output is chosen between RC buildspec and JSON, and then the output file is written. Is this high level understanding correct? Will look into the details more now, but from my understanding of the high level design it looks good!

Looks good to me!