THE FIELD GUIDE / AI & THROUGHPUT
AI providers, connections & feature routing
Choose named provider connections and compatible models independently for each AI workflow.
Set up a named connection
Configure and test independently scoped connections for providers such as OpenAI, Anthropic, Google Gemini or Vertex AI, Azure OpenAI and Amazon Bedrock. The provider catalog lists available adapters and capabilities. Multiple connections can use different credentials, regions or deployments within the same provider.
Route by feature
Choose a connection and model for scoring, process investigations, release planning and other AI-assisted workflows. Settings validates structured-output and evidence requirements before accepting a route. Runs retain the effective model and connection identity with redacted diagnostics.
Apply capacity and spending controls
Provider-specific limits and backoff combine with fleet-wide and per-server admission. Review pricing freshness, estimated and unsettled charges, and the impact of a route change on reusable evidence. A provider error can defer work while core inventory, patching and deterministic detection continue.
Open the application reference for capability scope and technical details.
Open View scope in issue #290 ↗