@retain_dev Yep! marpy parses your venture’s Python with AST and builds a reside, framework-aware map of it. Your SQLAlchemy/Django fashions (columns, number one keys, international keys, and relationships) and your FastAPI routers follows their Is dependent() chains.
That map is fed to the assistant on each and every request and to inline completions as you kind, ranked to no matter dossier you might be running in, so ideas are grounded on your exact cross-file schema and dependency wiring as an alternative of
guessed from a unmarried open dossier. It refreshes incrementally as you edit, so it will have to keep correct on larger initiatives.
At the SQLAlchemy migration ache particularly: marpy statically analyzes each and every migration and hard-blocks damaging operations (DROP/TRUNCATE) towards manufacturing. This can be a planned platform guardrail, now not an not obligatory atmosphere. So the “believable however broke” magnificence of recommendation is precisely what it is designed to catch.
