Moving one month is not the same as adding a fixed number of days. MonthAnchor keeps calendar code and an AI conversation together in Windows, helping you review clamped day values and supported year boundaries.
from calendar import monthrange
from datetime import date
def add_months(day, delta):
if type(day) is not date or type(delta) is not int:
raise ValueError("month.input")
year, month0 = divmod(day.year * 12 + day.month - 1 + delta, 12)
if not 1 <= year <= 9999:
raise ValueError("month.year")
last = monthrange(year, month0 + 1)[1]
return date(year, month0 + 1, min(day.day, last))from calendar import monthrange
from datetime import date
def add_months(day, delta):
if type(day) is not date or type(delta) is not int:
raise ValueError("month.input")
year, month0 = divmod(day.year * 12 + day.month - 1 + delta, 12)
if not 1 <= year <= 9999:
raise ValueError("month.year")
last = monthrange(year, month0 + 1)[1]
return date(year, month0 + 1, min(day.day, last))Date changes often conceal a product decision inside a utility function. MonthAnchor is a Windows AI coding workspace for discussing that decision with the implementation in view. Open your project and describe the expected calendar result before asking the connected model to edit the code. Include callers that depend on the current behavior, not only the date helper.
A recurring reminder offers a concrete task: moving a date forward one month when its day does not exist in the destination month. Ask the assistant to inspect the current rule and propose a change only after you decide whether the day should clamp or follow another policy. That reminder is a development scenario, not a supplied scheduling service.
Review the proposal across successive operations. Clamping once can affect what happens when another month is added, so define whether the original anchor day must be remembered elsewhere. Run project tests for month ends, leap years and the supported year range. Check the user-facing date as well as the helper’s immediate return value before accepting the edit.
MonthAnchor connects to supported OpenAI-compatible or Anthropic endpoints. An abliterated AI model can assist where offered by the configured service; context limits and tool support depend on that setup. The model can propose calendar code, while your requirements determine whether its behavior is correct.
The two operations answer different requirements. A reminder on a calendar date may not follow a fixed duration. Tell the assistant which rule the application promises, then inspect whether its proposed implementation preserves that rule for short months.
That depends on whether the application remembers an original anchor day or uses the last result as its new input. State the intended sequence explicitly. Test multiple moves rather than assuming a single correct result proves recurrence behavior.
Only if the affected requirement involves times or zone conversion. A date-only helper may have a deliberately narrower contract. Ask which callers need additional behavior before expanding the patch into a different class of calendar problem.
Include short destination months, a leap-year transition, negative movement and the supported year limits. Write the expected outcomes from your application policy. Run those fixtures in the project rather than relying on the assistant’s examples alone.
MonthAnchor · Updated 2026-10-25
