Expense Tracker

Log daily expenses via check-in prompts; dedup, categorize, sync to local ledger and Google Sheets.

Track daily spending via Telegram check-ins. Three prompts per day; cron-driven. All data flows to local .md ledger + Google Sheet.

Ledger format

Each file at ledger/YYYY-MM-DD.md:

markdown

# Ledger — 2026-06-03

- Food & Drinks $1.80 | Coffee
- AI $11.53 | Netlify

Format: - Category $Amount | Notes

Categories

Allowed: Food & Drinks | Groceries | Transportation | Household/Utilities | Pet | Shopping | Entertainment | Other | AI

• Bakery bread → Food & Drinks (NOT Groceries)
• Groceries = supermarket bulk/household runs
• Unsure → ask Hobo or default to Other with note

Check-in schedule

Time	Prompt
10am	Morning spend (coffee, breakfast, transport)
1:30pm	Afternoon spend (lunch, transport)
10pm	End-of-day recap

Each check-in runs as an isolated cron agentTurn with model: deepseek/deepseek-v4-flash and timeoutSeconds: 300.

Transport check-in (9am, previous day)

Separate cron asks: "Any transport yesterday?" Pinned to deepseek/deepseek-v4-flash.

Dedup rules

Before writing to today's ledger file:

1. Read today's ledger file first.
2. Time qualifiers (morning/afternoon/evening coffee): always ADD new entry, never replace.
3. Same category + amount: assume repeat purchase → ADD new entry.
4. Only REPLACE if Hobo explicitly says "change X to Y".
5. When in doubt → ask Hobo to confirm.

Expense sync

Daily at 9:30am via scripts/sync-ledger-to-sheet.py:

• Reads all ledger/*.md files
• Parses entries per format above
• Generates dedup key: (normalized_date|normalized_amount|notes) where normalize_amount() strips trailing zeros
• Compares against existing sheet rows → appends only new ones
• Uses Google Sheets API with service account at /home/hobopi/.openclaw/secrets/google-service-account.json
• Sheet ID: 1Ikbydh-Xzc6F3pk1Q5lbCTbdEERSb4Hq8obzbABVZbU, tab: Sheet1

Historical backfill

To backfill missing days (e.g., April 18-26 reconstructed from sheet data):

1. Query sheet for all rows with dates not in ledger
2. Parse each row as Date, Category, Amount, Notes
3. Create ledger/YYYY-MM-DD.md with entries in correct format
4. Verify count matches between ledger and sheet

Scripts

• scripts/sync-ledger-to-sheet.py — main sync engine
• scripts/gen-expense-data.py — dumps all entries to JSON for dashboard

Cron notes

• All expense crons MUST pin model to deepseek/deepseek-v4-flash (time-sensitive)
• Timeout: 300s minimum (600s for 10pm check-in)
• Use absolute paths only in payload messages — no cd && python3 chains