Finalize a target month — always refetch it fresh to confirm the movements are complete (all transactions present, no pending left), then recompile with --final to strip provisionals, and run heartbeat for the current month. Defaults to settling the previous month.
Assess a household's monthly investment snapshot for real anomalies. A script produces an objective month-over-month diff (value + quantity per holding, tagged qty/price/new/gone, plus integrity checks); then you reason over it and write a plain-language verdict — a summary message, an assess_{month}.md, and an assess_{month}.json (per-item, consumed by the viewer). Called at the end of /investments; also use whenever the user asks to review, sanity-check, validate, or "assess" a month's investments/portfolio/patrimônio, or whether anything looks unusual, missing, or wrong.
Build a household's monthly investment-portfolio snapshot by reading the broker statements (screenshots + PDFs) in investments/input and writing the canonical personal_finances_{YYYY-MM}.json. Use when the user asks to compile, snapshot, update, or generate investments / portfolio / patrimônio / posições for a month, or mentions BTG/Nubank/Wise/FGTS/Morgan Stanley statements under investments/input.
Validate the schema of a monthly budget, auto-fixing where possible. Checks required fields, types, buckets, FX, enrichment and installment rules. Use when /fetch, /recompile, or /heartbeat call it, or when the user asks to audit, validate, or check a month's data integrity.
Generate a monthly categorized budget JSON by running fetch + recompile, then classifying leftovers and forecasting partial months. Use when the user asks to generate, create, or analyze a monthly budget, expenses, or income.
Generate the two Telegram budget messages (leftover preamble + full advisory) from a compiled budget. Use when /compile, /heartbeat, or /settle need insights, or when the user asks for advice/analysis on their budget.
Fetch bank and credit card transactions from Pluggy API for a target month and write the raw month files used by the pipeline.
Fetch new Pluggy rows for the current month and recompile while preserving prior classifications. Use for the recurring mid-month update of the open month.