OpenAI 额度锁死应急：配置回退 provider，避免会话连续 `failed before reply`
解决 OpenAI 余额归零后网关进入 cooldown、会话持续失败的问题。前置：至少两个可用 provider，且可调整默认模型策略。步骤：识别账单锁定→切换/配置回退 provider→重启并复测关键链路→恢复后再切回。关键命令：`openclaw gateway status`、`openclaw gateway restart`、`openclaw status`。验证：账单异常期间仍可由回退模型稳定回复。风险：若 memory/embedding 仍绑定失效 provider，部分功能可能继续降级（需验证）。
GITHUBDiscovered 2026-02-15Author EmpireCreator
Prerequisites
- At least one fallback provider is pre-configured and authenticated.
- You can inspect gateway error logs for quota/billing/cooldown signals.
Steps
- Confirm the failure signature (`quota exceeded`, provider cooldown, repeated pre-reply failure).
- Switch default model/profile to a healthy fallback provider and keep prompts unchanged.
- Restart gateway and run smoke tests for chat, tool calls, and memory lookup paths.
- After billing recovery, re-enable primary provider gradually and monitor for cooldown recurrence.
Commands
openclaw gateway statusopenclaw gateway restartopenclaw statusVerify
During OpenAI billing lock, session stays serviceable via fallback with no repeated user-facing failure loops.
Caveats
- Do not silently hide outages; surface one actionable incident message to operators.
- Provider latency/cost profile may change after failover（需验证）.
Source attribution
This tip is aggregated from community/public sources and preserved with attribution.
Open original source ↗