Description: When configuring a custom model under BYOK (Bring Your Own Key), I explicitly set the Context Window to 512,000 and the Max Output to 128,000.
However, during active sessions in the Cuppy chat window, Cuppy ignores the user-defined context window limit and instead defaults to the model provider’s official maximum context window.
Expected Behavior: Cuppy should respect the custom context window configured in BYOK and automatically trigger context compression/pruning once the conversation reaches the user-defined limit (512,000 tokens).
Actual Behavior: Cuppy bypasses the custom threshold and allows the context to expand up to the model provider’s official maximum default without compressing.
Impact: Because upstream model providers utilize tiered pricing based on active context length (e.g., higher rates for larger token tiers), ignoring the custom context limit results in significant, unexpected API expenses.