Use case overview
This group of pages is organized by "what you want to achieve," tying the gateway's relevant capabilities into actionable solutions. Each page points to specific how-to guides and configuration references — they are not dead-end pages.
Not sure where to start? Go back to the solutions-by-goal table on the documentation home page, or first get Docker single node running and then branch by goal.
| Use case | For | Difficulty | What it solves |
|---|---|---|---|
| Enhance model capability | Administrator | Intermediate | Fill model gaps: auto-switch image-bearing requests to a multimodal model, swap in a cost-effective model during Claude Code's execution phase, multi-model consensus for PR code review |
| Reduce call cost | Administrator | Intermediate | Use upstream prompt caching, model tiered routing, response trimming, and log retention policy to cut gateway-side call cost |
| Achieve high availability | Administrator | Intermediate | Use a load balancer for multi-node failover, provider-level disaster recovery, multi-instance shared state, and canary shutdown |
| Multi-tenant isolation | Administrator | Advanced | Use key groups to split models and load balancers, use request-header rules to distinguish clients, per-tenant rate limiting and billing |
| Office / Microsoft 365 add-in | Administrator | Advanced | Self-host the Office add-in on your own origin so users connect directly to your own gateway inside Excel / Word / PPT / Outlook, with zero dependency on pivot |
| Meet compliance and audit requirements | Administrator | Advanced | Request and response retention, two-layer sensitive-information redaction, login and permission audit, retention period and deletion |
| Troubleshoot from symptoms | Integrator + Administrator | Intermediate | Use z-request-id to locate 401 / 403 / 404 / 502 / 503 / timeout / streaming interruption / memory anomaly symptoms |