Review before applying
Compare actual behavior with the expected result. If a check fails, investigate it before applying more changes.
AI coding workspace
Bézout integer coefficients
BezoutVale is a Windows AI coding client for reviewing integer algorithms. Discuss coefficient signs through a compatible endpoint and check the proposed code against the defining identity before accepting changes.
● ● ●BezoutVale●
A concrete starting point · Bézout integer coefficients ×
def bezout(a, b):
if any(type(v) is not int or abs(v) >= 10**100 for v in (a, b)):
raise ValueError("integer.range")
old_r, r = abs(a), abs(b)
old_s, s, old_t, t = 1, 0, 0, 1
while r:
q = old_r // r
old_r, r = r, old_r - q * r
old_s, s = s, old_s - q * s
old_t, t = t, old_t - q * t
return old_r, old_s * (-1 if a < 0 else 1), old_t * (-1 if b < 0 else 1)✳Return nonnegative gcd and Bézout coefficients for two integers with fewer than 101 decimal digits.
✓ ChangesReview before applying ↗
Illustrative workspace
Connection
Base URL, credentials and a reachable server.
Project
Test negative operands and both zero cases, verifying the integer identity and rejecting Boolean inputs.
A concrete starting point
def bezout(a, b):
if any(type(v) is not int or abs(v) >= 10**100 for v in (a, b)):
raise ValueError("integer.range")
old_r, r = abs(a), abs(b)
old_s, s, old_t, t = 1, 0, 0, 1
while r:
q = old_r // r
old_r, r = r, old_r - q * r
old_s, s = s, old_s - q * s
old_t, t = t, old_t - q * t
return old_r, old_s * (-1 if a < 0 else 1), old_t * (-1 if b < 0 else 1)Review the return contract as carefully as the calculation
BezoutVale is Windows coding software for examining helper routines whose outputs have to satisfy a precise contract. Open the project and describe what the caller expects from each returned value. Ask the model to inspect that interface before proposing a refactor, especially when several outputs can appear plausible in isolation.
A task based on bezout.py could focus on signed inputs and the coefficients returned with a common divisor. Bring a negative-input fixture into the conversation and state the identity the result must satisfy. Then review how the caller consumes the tuple, so a reordered return or changed sign convention does not pass unnoticed behind correct-looking arithmetic.
Use independent checks to evaluate the proposal. Run the project’s tests and inspect zero-input combinations, sign variations and rejection of unsupported operands. Keep the evidence attached to the original requirement. A model’s explanation can guide the review, but the application needs results that satisfy both the mathematical relation and the interface it actually calls.
Choose your model through a supported OpenAI-compatible or Anthropic endpoint, supplying credentials as required. An uncensored AI model may be available from that service, subject to provider restrictions. Tool support and context capacity depend on the selected model and server; scope the source discussion to what can be examined reliably in the session.
Questions for a multi-value arithmetic interface
Must two correct implementations return identical coefficients?
Not necessarily. Establish which properties the application requires and whether it relies on one particular representation. An equality assertion over a whole tuple may be stricter than the contract. Review the caller before changing either the routine or its tests.
Why inspect negative inputs separately?
Sign handling can preserve a plausible divisor while breaking the relation involving the coefficients. Provide signed fixtures and verify the defining relation independently. Review any normalization at the input boundary so the returned values still refer to the original operands.
What should happen when both inputs are zero?
State the intended result or rejection policy and compare it with the current implementation and callers. Do not assume ordinary positive-input reasoning settles the zero case. Include it explicitly in the review so later refactors preserve the chosen contract.
Can I request a cleanup without changing return conventions?
Yes. Put the existing tuple order, accepted types and sign rules in the request, then ask for a bounded change. Inspect the resulting diff for altered validation or reordered outputs and run contract-level checks before accepting the cleanup into your project.
BezoutVale · Updated 2026-10-25