Streaming responses now preserve quota and usage accounting by recording completion metadata after SSE stream closure.
By QuizMind Platform Team
stream=true requests are now first-class in the proxy controller.
Usage extraction from SSE events updates quota snapshots consistently.
Completion persistence still runs even when streams close on the client side.
The proxy controller now bridges readable web streams to the HTTP response pipeline with controlled abort handling.
The service inspects SSE chunks for usage and response identifiers, then records completion metadata after stream consumption.
Quota counters and usage reports stay accurate for streaming and non-streaming workloads.
This keeps billing, abuse detection, and operator analytics aligned with actual runtime traffic.