@thsottiaux Codex GPT models keep forcibly redacting my debug logs, even when I explicitly disable it in the rules. When errors occur, all critical values are masked, making the issue impossible to locate. The model then has to search the codebase and add new logs. Please fix this.
@thsottiaux The root cause is Codex GPT models’ obsession with redacting all kinds of information. They sanitize debug logs even when rules forbid it, masking the exact values needed to diagnose errors. This makes debugging much slower and forces the model to add new logs. Please fix this.
@thsottiaux During peak hours: Reduce inference speed for users who don't use Fast mode; increase the usage limit for existing users who do use Fast mode to ensure speed meets the needs of users. Off-peak hours: Increase usage limit and reduce consumption speed.