SSE and WebSocket continuity
Inspect sustained delivery, idle periods, disconnects and reconnections against controlled endpoints. This helps investigate interrupted streams but does not validate an AI login, quota or model request.
Choose a duration and idle scenario
Choose 30, 60 or 120 seconds, optionally set a 0, 5 or 15 second idle interval, and decide whether to include one planned server disconnect. Observe the actual stages, message counts and event times for SSE and WebSocket.
Stages come from connection events and server frames, not estimated completion percentages. Setup has an additional eight-second budget and transport completion fifteen seconds. Stop at any time; leaving the page also cancels the check.
Check completion, interruptions and recovery
A completion frame, continuous sequence, recovery after planned idle and successful planned reconnection together establish complete flow evidence. The longest gap includes planned idle; sequence gaps are application observations, not UDP loss.
Teaching example: continuous delivery works, idle recovery fails
Save A with continuous delivery, then B with planned idle on the same network. If B fails, review proxy idle timeouts and exclude device sleep or background throttling. This example does not claim a fault was detected in your proxy.
Does success validate an AI service?
This tool targets this site’s controlled endpoints and covers only the path from your browser to this site. ChatGPT, Claude and other platforms use different targets. Accounts, quotas, regional eligibility and model calls still require verification on the relevant platform.
Related tools and guides
Live measurements run in your browser after JavaScript loads. This introduction contains no visitor detection results.