Jeroen UbbinkandClaude Fable 5 3f850c54f4 fix: re-reserve the serving candidate's max cost on failover ...

Admission and reservation were sized once to the best-ranked candidate's
max cost while settlement bills the candidate that actually serves, so a
failover to a pricier candidate could settle far beyond the admitted
envelope and consume balance reserved by other in-flight requests. Before
trying a fallback candidate, raise the reservation to its own envelope;
reject candidates the key cannot cover, exactly as admission would have
had they been ranked first.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-07-22 15:12:21 +02:00

..

integration

fix: re-reserve the serving candidate's max cost on failover

2026-07-22 15:12:21 +02:00

unit

fix: fail over with each candidate provider's own model

2026-07-22 15:12:21 +02:00

run_integration.py

dump wip stash

2025-08-06 20:31:55 -03:00