classify

name	classify
description	Classify specific transactions (expenses or income) using the household's expenses_memory.md and income_memory.md. Receives 1 or more transactions via reply or inline. Updates memory files with new patterns. Use when the user asks to classify specific transactions — NOT for whole-month batch (that's /categorize).

Classify Transactions

Classifies specific transactions (1 or more) — not a whole month. For whole-month batch classification, use /categorize.

Input

The user provides specific transactions to classify, typically by replying to a bot message that contains them, along with classification instructions (e.g. /classify lucas andrade as services designer).

• The replied-to message text — contains the transaction(s) to classify
• The user's instructions — may specify category, subcategory, type, or other hints

Process

1. Parse transactions from the replied message or inline input
2. Apply the user's instructions for classification
3. Match against expenses_memory.md and income_memory.md for context
4. Update expenses_memory.md with the new pattern so future occurrences are auto-classified

Output

• Confirmation of the classification applied
• For each transaction: type (expense/income), category, subcategory, bank

Reference Files

• resources/{household}/expenses_memory.md — Known Merchants table, Manual Overrides, category hierarchy, budget bucket mappings
• resources/{household}/income_memory.md — Salary definitions, known income sources, date windows, frequencies

Classification Logic

For each transaction:

1. Determine type: expense or income

• Match against income_memory.md salary rules (amount range + date window) and known income patterns
• If it matches an income pattern → income
• Otherwise → expense

2. Assign category and subcategory

• Expenses: match description against expenses_memory.md merchant patterns (case-insensitive, partial match). Manual Overrides take precedence over Known Merchants. Auto-classify unmatched by name heuristics (e.g. "Posto" → Transportation/Fuel, "Drogaria" → Health/Pharmacy)
• Income: match against income_memory.md definitions and known patterns (salary, cashback, IOF adjustment, FGTS, named transfers)

3. Update description when opaque

• If the original description is generic or opaque (e.g. "Transferência Recebida", "Pagamento efetuado"), append a clarifying suffix in parentheses with the classification context (e.g. "Transferência Recebida (FGTS Saque Aniversário - holder2)").
• Do not replace the original text — only append. This preserves traceability to the bank statement.

4. Flag unknowns

• If no match is found, flag as Uncategorized so the user can review

Special Handling

Transaction Type	Treatment
Aplicacao RDB (small auto)	Investment / Troco Turbo
Aplicacao RDB (large, intentional)	Skip — not an expense
Estorno / refund	Allocate to original merchant's category as negative amount
Reimbursements (e.g. CARE PLUS)	Health / Reimbursement — net against health expenses
IOF de compra internacional	Income / Adjustment (IOF reimbursement)
Skip-flagged in expenses_memory.md	Excluded from totals

Memory Update

After classifying, run /learn to persist any newly discovered patterns to memory files. This replaces manual memory updates — /learn handles adding new merchants and income sources automatically.