What's Changed
New: tally explain Command - AI-Friendly Classification Reasoning
Ever wonder why a merchant ended up in "Variable" instead of "Monthly"? Now your AI agent can find out.
The new tally explain command exposes the full decision-making process, making it easy for LLMs to understand classifications and suggest better rules:
tally explain # Overview of all classifications
tally explain Netflix # Explain specific merchant
tally explain Netflix -vv # Full details + which rule matched
tally explain --category Food # Explain all Food merchantsExample output with -vv:
Netflix → Monthly
Monthly: Subscriptions appears 6/6 months (50% threshold = 3)
Decision trace:
✗ NOT excluded: Subscriptions not in [Transfers, Cash, Income]
✓ IS monthly: Subscriptions with 6/6 months (>= 3 bill threshold)
Calculation: avg (CV=0.00, payments are consistent)
Rule: NETFLIX.* (user)
The Rule: line shows exactly which pattern matched and whether it came from your merchant_categories.csv (user) or the built-in rules (baseline) — so agents can confidently suggest rule additions or modifications.
New: JSON/Markdown Output Formats
Export analysis with full reasoning data for programmatic consumption:
tally run --format json -v # JSON with decision traces
tally run --format markdown # Markdown report
tally run --only monthly,travel # Filter to specific classificationsPerfect for piping into your AI agent's context or building custom dashboards.
Improved Error Messages
CLI errors now guide you to the right answer:
- Invalid
--onlyvalues warn and show valid options - Invalid
--categoryshows available categories - Typos in merchant names suggest similar matches (fuzzy matching)
Install
Linux / macOS:
curl -fsSL https://tallyai.money/install.sh | bashWindows PowerShell:
irm https://tallyai.money/install.ps1 | iexOr download the zip for your platform from the assets below.
See https://tallyai.money for more info.