示例场景 / Example Use Cases

Multiple provider accounts made budget reviews slow. Long prompts (RAG) needed clearer routing rules, and the finance team wanted one place to review usage.

多个供应商账户让预算复盘变慢。长 prompt (RAG) 需要更清楚的路由规则，财务团队也希望在一个地方查看用量。

Switched to one thistoken.ai gateway key. Migrated production code in 1 day by changing baseURL, then set route policies for short queries and long-context work.

切换到 thistoken.ai 网关 Key。1 天内通过修改 baseURL 完成迁移，并为短查询和长上下文任务配置路由策略。

Agent loops with 10-50 round trips per task needed better model selection. A single-model strategy made it hard to balance quality, latency, and budget.

Agent 循环每个任务 10-50 次往返，需要更好的模型选择。单一模型策略很难同时平衡质量、延迟和预算。

Adopted tiered routing: classification and simple steps use one policy, reasoning uses another, and final synthesis keeps a higher-quality model. All remain behind the same API.

采用分级路由：分类和简单步骤使用一个策略，推理使用另一个策略，最终合成保留更高质量模型。全部保持在同一个 API 下。

Customer-facing workflows needed clearer fallback rules, incident visibility, and policy controls for sensitive requests.

面向用户的支持流程需要更清楚的兜底规则、事件可见性，以及敏感请求的策略控制。

thistoken.ai as the gateway layer, with automatic failover when latency crosses the project threshold or a sensitive intent policy is triggered.

以 thistoken.ai 作为网关层，当延迟超过项目阈值或触发敏感意图策略时自动切换到兜底路由。