Can a prompt replace Tiller?

Build a local, spreadsheet-shaped budgeting ledger that imports bank transaction exports rather than connecting live to bank accounts — no screen-scraping or third-party aggregation, by design. Use a Next.js/TypeScript app with SQLite, or, if you'd rather stay closer to Tiller's actual model, a Google Sheets API integration that writes into a spreadsheet template you own — either is a legitimate approach. Support CSV and OFX import with a saved column-mapping per bank (so re-importing the same bank's export doesn't require re-mapping every time), and duplicate-transaction detection by date, amount, and description matching. Implement accounts, categories, payees, and split transactions (one transaction covering multiple categories). Add deterministic categorization rules — if description contains X, categorize as Y — that are visible and editable, not a black-box classifier, so every categorization can be traced back to the rule that produced it. Build budget, cash-flow, and net-worth views computed directly from the ledger. Keep an immutable import log so you can always see exactly what was imported and when. Do not build live bank-account connections, payment initiation, or tax/accounting compliance features — those are explicitly out of scope; this is an import-and-categorize ledger, not a bank aggregator. No API key needed for the CSV/OFX path; the Sheets API path needs a free Google Cloud project credential.