Solutions / Catalogue / Acoustic & Audio AI

Acoustic & Audio AIReference design

Voice-Controlled Warehouse Management

A reference build where a warehouse worker speaks a stock movement and it is written straight into inventory through MCP tools. Five tiers go from a cloud trial to a fully on-premise setup. It fits receiving docks, back rooms, tool cribs and line-side stores where hands are busy.

LogisticsManufacturing

Talk to an engineer See the hardware

Two-level warehouse with tall shelving aisles stocked with cartons, a worker walking between racks — Photo: Adam Hall GmbH, CC BY-SA 3.0 de, Wikimedia Commons. Shows the setting, not this installation.

about 16 tokens/s: local LLM speed on J5011 (vendor measurement)
14.8 ms: median material list query time with 10 concurrent requests on R2000 Series
22.1 ms: P99 material list query time on the same test
600/min: stock transaction rate limit per user

Results reported by the design's authors on their own test setups. Yours depend on the site, so we confirm them in a pilot.

The problem

Why it matters.

Warehouse staff with full hands cannot type, so stock moves are logged late or not at all.

The flow

From signal to action.

01PeopleSpeakThe operator wakes the SenseCAP Watcher and says the movement, such as items received.
02DecideSpeech AISpeech is recognised and parsed in SenseCraft cloud, on a J4012, or fully on a J5011.
03ConnectMCP callThe speech AI calls the matching inventory tool on the gateway.
04RecordGateway writeA Docker service on a reComputer saves the record in SQLite.
05ActConfirmThe Watcher speaks and shows the new stock total.
06PeopleDashboardThe movement appears with operator and reason, and ERP or WMS can read it.

Deployment architecture

How it is wired.

The devices on site, what carries their data, where the decision is made and where the result goes.

01SenseOn site

SenseCAP Watcher

02DecideRuns the model

reComputer R1125-101 per site
reComputer Industrial R2135-121 per site
reComputer Super J40121 per 3 Watchers
reComputer Robotics J50111 per site
SenseCraft
EdgeLLM

03Your systemsWhere results land

MCP
REST API
SQLite
ERP
WMS

04ActWhat happens next

reTerminal D1001
Confirm

How it works

Step by step.

01Pick a tier by how much must stay on site and whether you need face ID.
02Flash the Watcher over USB-C and join it to 2.4 GHz Wi-Fi.
03Deploy the gateway service and import stock from the Excel template, or map your ERP to the MCP tools.
04Enrol operator faces if each entry should carry a name.
05Staff speak short transactions and hear the result.
06Check the dashboard and export records as needed.

The hardware

What goes on site.

Items we stock can go straight in the bag. Everything else we source and quote with the installation.

SenseCAP Watchervoice terminal with on-device face ID1 per work point, 3 in the multi-site tierSourced on quote
reTerminal D10018 inch touch alternative to the WatcherSourced on quote
reComputer R1125-10gateway holding inventory data (Tier 1)1 per siteSourced on quote
reComputer Industrial R2135-12gateway with Hailo-8 face ID and liveness (Tier 2A and 3)1 per siteSourced on quote
reComputer Super J4012local speech, warehouse system and face ID for up to 3 Watchers1 per 3 Watchers₹1,59,389
reComputer Robotics J5011fully on-premise LLM and speech (Tier 3)1 per siteSourced on quote

Before you install

2.4 GHz Wi-Fi at the work points; 5 GHz-only networks will not show up.
Internet for the Trial, Tier 1 and 2A, and for LLM calls in 2B; LAN only for Tier 3.
About 10 GB free disk on the J5011 for Tier 3 models.
An ERP or WMS with a writable interface, or use of the built-in system.

Good to know

There is no offline queue; writes fail during a network outage and must be repeated.
Not suited to long batch or serial codes, multi-step dialogues or very noisy areas.
On-device face ID is limited to 20 operators.
The Tier 3 local model is fixed at Qwen3.5-4B.

How we deliver it

Paper to site, one team.

01Site surveyWe walk the site, or review photos and drawings, and agree what to measure and where each device goes.
02PilotOne lane, one line or one room first. We tune it on your conditions and agree what good looks like.
03Roll outWe supply the hardware in India with GST invoices, install it, and connect it to the systems you already run.
04SupportSpares, remote checks and model updates, with one team to call when something changes.

Questions

Asked before buying.

Can it run fully offline?

Yes in Tier 3, which runs speech and the LLM on a J5011 with no outside calls after setup.

How hard is it to connect our existing ERP?

You map the ERP interface onto the gateway's MCP tools. The page describes this as a small amount of mapping code.

What happens if Wi-Fi drops mid-entry?

The write fails and is not queued, so the operator repeats it once connected.

Want this on your site?

Tell us the site and the problem. We come back with a pilot plan and a priced hardware list.

Book a call with an engineer Write to us

Want this on your site?

In-Person Conversation Intelligence

Counter Audio Capture & Analytics

AI Voice Warehouse Management

Want this on your site?

In-Person Conversation Intelligence

Counter Audio Capture & Analytics

AI Voice Warehouse Management