TrackLLM0006

ChatGPT Style Chat Service

Hard

Problem

Serve a chat product to 100K concurrent users with streaming responses and first token under 1 second at p95. Under load spikes it degrades gracefully. The GPU count must be justified by arithmetic.

Required components

Click to add a component to the canvas.

Additional components

Browse by category. Click or drag a component onto the canvas.

Drag components from the library onto the canvas, then drag from a component's right handle to connect it to another.

100%

Design canvas

Sign up to start designing

Sign up free

Premium is free for 83 more days.