Problem
azd ai agent invoke has a --long-running option (and optional --no-wait), but extension usage telemetry does not distinguish long-running invocations from regular remote invokes. We cannot measure adoption or compare the two modes.
Proposed change
Emit one best-effort agent.invoke.selected extension usage event after resolving a remote invoke target and before the invoke request. Project-backed routes must resolve a hosted agent; explicit --agent-endpoint routes have no project kind to inspect. Resolution may authenticate or probe a brownfield agent first. Include only:
agent.invoke.protocol: the resolved invocable protocolagent.invoke.long_running: whether --long-running was selectedagent.invoke.no_wait: whether --no-wait was selected
This measures selected remote routes, not all attempts, service acceptance, or success. Exclude local and non-hosted project invokes; do not emit names, IDs, endpoints, prompts, or raw errors. Keep the event schema, public telemetry reference, coverage matrix, and tests in sync.
Implementation: #10236
Problem
azd ai agent invokehas a--long-runningoption (and optional--no-wait), but extension usage telemetry does not distinguish long-running invocations from regular remote invokes. We cannot measure adoption or compare the two modes.Proposed change
Emit one best-effort
agent.invoke.selectedextension usage event after resolving a remote invoke target and before the invoke request. Project-backed routes must resolve a hosted agent; explicit--agent-endpointroutes have no project kind to inspect. Resolution may authenticate or probe a brownfield agent first. Include only:agent.invoke.protocol: the resolved invocable protocolagent.invoke.long_running: whether--long-runningwas selectedagent.invoke.no_wait: whether--no-waitwas selectedThis measures selected remote routes, not all attempts, service acceptance, or success. Exclude local and non-hosted project invokes; do not emit names, IDs, endpoints, prompts, or raw errors. Keep the event schema, public telemetry reference, coverage matrix, and tests in sync.
Implementation: #10236