Use Verdictan
Task Guides
Pick the task you want to do. Every guide ends with a check that proves the result.
New to Verdictan? Send your first request before you set up an integration or change a policy.
Connect an application
Send your first requestConnect a local gateway from the Console and see the response, decision, and usage.Start with DockerRun the gateway in a container with the supplied Compose file.Connect a model providerGive the gateway access to your provider account and test a model request.Connect an editorPoint the supported client of your editor at the gateway.Connect a Node.js appReuse the OpenAI-compatible client of your application.Connect a Python appReuse the OpenAI-compatible client of your application.
Control requests and cost
Configuration in a fileEdit the gateway configuration in Git and test it before deployment.Start from a templateAdapt a starter configuration and test its decisions before rollout.Change a policyTest a small change, roll it out to a canary, and keep a way back.Review usage and budgetsFind what consumes your budget and choose a spending control.Compare two modelsCompare quality, cost, and latency for the same task.
Manage a team
Prepare your teamAgree who owns access, configuration, request review, and spending.Invite membersAdd people and organize them into teams.Assign permissionsGive a role only the actions its members need, and test it.Set up company sign-inConnect your identity provider and test sign-in before you enforce it.
Understand an unexpected result
Investigate a blocked requestFind the decision event and the policy that blocked it.Resolve an escalationReview a held request and record a decision in the Inbox.Review an alertCollect the request, configuration, and audit evidence.Find who changed a resourceRead the change record in Trail and verify its integrity.Export evidenceDownload decision events and verify the file digest before handoff.Fix a failed setupStart from the error you see and check the connection behind it.
A healthy gateway only shows that its process runs. A correct model response also shows that your application authenticates and reaches the provider. Use the check at the end of each guide to prove the behavior you changed.