Since we have the streaming cancellation propagation issue, we can not rely on the Cody Gateway per model latencies to understand the E2E performance. To clarify the exact model used for every completion, we can use the x-cody-resolved-model header and add this information to client-side logs.
Adds x-cody-resolved-model as resolvedModel to autocomplete analytics events.
Since we have the streaming cancellation propagation issue, we can not rely on the Cody Gateway per model latencies to understand the E2E performance. To clarify the exact model used for every completion, we can use the
x-cody-resolved-modelheader and add this information to client-side logs.x-cody-resolved-modelasresolvedModelto autocomplete analytics events.Test plan
Updated unit tests. To verify manually:
gatewayModelbeing logged along with theCompletionLogger:onCompleteevent.