Step 1
Map the workloads that will send supported Chat Completions, Responses and Messages traffic through the gateway.
Platform · Routing
Route every supported request to an approved model target
Platform teams need explicit control over which provider and model handle each workload. Verdictan routes supported gateway traffic to configured upstream targets and keeps credentials scoped to those targets.
What you are trying to solve
Ad hoc model switching in application code can bypass spend controls and policy tests tied to a specific target.
Step 2
Configure the product area that matches your priority: policy, routing, spend, agents, performance, or operations.
Step 3
Run deterministic policy tests and review decision samples before you expand routing.
Routing decisions teams document
Separate development, staging, and production targets in configuration so experiments do not share credentials with customer-facing workloads.
Routing applies to supported request families only. Verify compatibility for each integration before rollout.
Capabilities teams configure first
Target allow lists
Keep production traffic on approved providers and model IDs defined in configuration.
Environment separation
Use distinct routing tables per environment to prevent credential bleed.
Evidence of selection
Retain gateway events that show which target handled each supported request.
Rollout sequence
- Inventory provider targets already in use across services.
- Move supported traffic to the gateway and map each workload to an approved target.
- Review routing changes through the same policy test process you use for control updates.
Related paths
Continue with a template, product area, or guide that matches this page.
Application teams keep a stable gateway endpoint while platform teams control targets centrally.
Download Verdictan and validate the control path in your environment.