Your voice agent sounds right. Did it place the right order?

Описание и применение

IndicOrderBench выявил несоответствие между заявленным и фактическим заказом в голосовом агенте, предотвращая ошибку до взаимодействия с клиентом.

Задача

Voice ordering agents may misinterpret or miscommit orders, especially with code‑mixed speech, leading to incorrect backend state.

Результат

IndicOrderBench detected the mismatch between committed and expected order, causing CI failure before customer impact.

Рабочий процесс

• Customer speaks request in Hinglish
• Agent parses and confirms order
• Agent commits order to backend
• Backend state is compared to expected state

Доказательства

The agent read back one wrap without onion. This is what it committed:

Источник

A transcript-based test, or an LLM judge reading the conversation, passes this call. The customer gets two wraps with onion.

Источник

IndicOrderBench is an open-source benchmark and CI gate for voice ordering agents.

Источник