AI WebGPU Lab Experiment

LLM Worker UX Readiness

`exp-llm-worker-ux` now exposes a worker/main comparison harness for local chat UX. It keeps prompt budget stable while recording TTFT, decode throughput, turn latency, and a lightweight responsiveness probe.

Use `?mode=main` to compare the main-thread path against the default dedicated worker path.

Prompt

Responsiveness Probe

Run Controls

Run one chat turn to compare the active execution mode against the stable worker/main result contract.

Output Stream

No chat turn yet.

Metrics

Environment

Activity Log

    Schema-Aligned Result Draft

    {
  "status": "pending"
}

    Mode Notes

    • Default mode uses a dedicated worker for the simulated chat turn.
    • `?mode=main` keeps the same prompt/output budget but records main-thread execution metadata.
    • The responsiveness probe is summarized in `meta.notes` until a dedicated UX metric schema is added.