Real demo recording · English narration and captions · Fictional data
THE PROBLEM
An operator needs device information without losing visibility into the data the assistant uses.
THE WORKFLOW
Ask about a fictional inventory, inspect the device reading and play a spoken answer. Tools are read-only.
01Voice question
02Bounded intent
03Inventory lookup
04Inspectable answer
RECORDED EVIDENCE
Results, with context.
Results on synthetic data
Check
Cases
Device selection
12/12
Intent identified
12/12
Dataset
synthetic-regression-v3-english
Completed cases
12/12
Median / p95 per case
0.004 s / 0.638 s
Evaluation record (UTC)
2026-10-08
Source: the project’s evaluation record, using local inference. Recorded hardware: NVIDIA GeForce RTX 5060 Ti, 595.84, 16311 MiB. Latency includes the full HTTP case and may include initial model loading. These measurements have not been rerun on the machine serving this website.
Evaluated demo