Supported Hardware¶
Pantry Raider runs as a set of Docker containers, so it works on most 64-bit Linux hardware. This page lists what's officially tested and the minimum specs to expect a good experience.
Looking to flash a ready-to-go image instead of installing manually? See the SD-card image guide.
Planning which accessories to buy for a Pi build? The Pin map and accessory compatibility page covers power, PoE HATs, plug-in sensors, and sound, and says which combinations work together.
Minimum requirements¶
| Resource | Minimum | Recommended |
|---|---|---|
| Architecture | ARM64 (aarch64) or x86-64 | 64-bit only |
| RAM | 2 GB | 4 GB |
| Storage | 16 GB | 32 GB+ (Mealie recipe images add up) |
| OS | 64-bit Linux with Docker + Compose v2 | Raspberry Pi OS Lite (64-bit) / Debian / Ubuntu Server |
| Network | Ethernet or Wi-Fi | Ethernet for the always-on box |
RAM guidance: Pantry Raider + Grocy run comfortably in 2 GB. Adding Mealie (recipes/meal plan/shopping) pushes the practical floor to 4 GB, especially during meal planning, so an install on a 2 GB board leaves Mealie off by default (you can turn it on later on a bigger board). Local AI via Ollama is not recommended on low-RAM SBCs: use a cloud AI provider, or a machine with 16 GB+ if you want fully local inference. For the full per-board stack recommendations and the SD-card protections a Pi install applies, see Pi reliability and memory tiers.
Tested boards¶
| Board | RAM | Status | Notes |
|---|---|---|---|
| Raspberry Pi 5 | 4 GB / 8 GB | ✅ Supported | Recommended. DSI touch display support. |
| Raspberry Pi 4B | 4 GB / 8 GB | ✅ Supported | Solid; 2 GB works for Grocy-only setups. |
| Raspberry Pi 4B | 2 GB | 🟡 Limited | Pantry Raider + Grocy only. Mealie may be tight. |
| Generic x86-64 mini PC (N100, etc.) | 8 GB+ | ✅ Supported | Runs everything; best for local Ollama. |
| Raspberry Pi 3B+ | 1 GB | 🟡 Pi Remote only | Insufficient RAM for the full stack; fine as a thin Pi Remote kiosk or Stream Deck. |
| Raspberry Pi Zero 2 W | 512 MB | ❌ Unsupported | Insufficient RAM. |
Status key: ✅ Supported (tested, expected to work well) · 🟡 Limited (works with caveats) · ❌ Unsupported (known inadequate).
This matrix is filled in from real testing on the SD-card image. If you've run Pantry Raider on hardware not listed here, please open an issue with your results.
Peripherals¶
Barcode scanners¶
Any USB HID ("keyboard wedge") barcode scanner works with no configuration: it types the scanned code into the focused field. This covers most wired and wireless USB scanners, including compact OEM scan-engine modules. 1D (UPC/EAN) and 2D (QR/DataMatrix) are both supported as long as the scanner reads them.
For a fixed kiosk you can run a scan-engine module hands-free (scan on sight, no button). The Waveshare Barcode Scanner Module guide has the ready-to-scan configuration codes for that.
On a Raspberry Pi appliance you can also wire a scan engine straight to the serial pins, no USB port used. A 3.3V Waveshare-style engine wires directly and stays dark until a scan page is open. The SEENGREAT Barcode Scanner Reader is also supported, with two caveats: it is a 5V module, so its data-out wire needs a level shifter or a two-resistor divider before the Pi (the Pi's serial input is not 5V tolerant), and it ships acting as a USB keyboard, so you must scan its "serial port output" and "continuous scanning" configuration barcodes once before the serial pins say anything. The wired UART barcode scanner guide has the wiring and the one-time setup for both.
The camera-based scanner in the web UI also works on any device with a camera (e.g. your phone), no dedicated hardware required.
Displays¶
A display is optional: Pantry Raider is a web app you can reach from any browser on your network. For a dedicated touchscreen setup, DSI and HDMI capacitive touch panels both work; see the SD-card image guide for kiosk-mode setup.
Recommended minimum size: 7 inches for full functionality. Smaller displays are supported, but not every page fits comfortably and some controls get cramped, so a small panel is best used as a simple info display (timers, weather, expiring items) or paired with a Stream Deck controller that handles the navigation while the screen shows content.
Stream Deck controllers¶
An Elgato Stream Deck, or an embedded Stream Deck Module, can act as a physical controller. Its keys show live counts (items expiring soon, scans waiting to commit) and trigger actions such as committing pending scans or opening a page on the attached display. It can sit next to a touchscreen or be the only interface on a headless box.
| Device | Keys | Status | Notes |
|---|---|---|---|
| Stream Deck Mini / Module 6 | 6 | Supported | Extra actions move to further pages via a "More" key. |
| Stream Deck / MK.2 / Module 15 | 15 | Supported | Roomy default layout. |
| Stream Deck XL / Module 32 | 32 | Supported | Plenty of spare keys. |
Setup, configuration, and the controller service live in
streamdeck/. The connection is plain USB and
the driver is pure Python, so no Elgato software is involved. Use a
data-capable USB cable (charge-only cables leave the deck lit but undetected)
and a full-strength power supply; see
Power and cabling.
Cameras / photos¶
No dedicated camera is needed. Use your phone's browser to photograph food items and receipts, the native camera opens directly from the web UI.
Presence sensors¶
A Pi appliance with a display can wake the screen as you walk up, using a 24GHz mmWave presence sensor on the GPIO header. The tested part is the HLK-LD2410C (about $5); see mmWave presence sensor for which modules work and how to wire them.
Bluetooth kitchen thermometers¶
A Pi appliance with a Bluetooth radio (built in on a Pi 4/5) can read meat and probe thermometers directly: Inkbird, ThermoPro (including the TempSpike), Combustion Inc, ThermoWorks BlueDOT, and Govee grill thermometers. A server with no Bluetooth radio can still read the same thermometers through Home Assistant. See Bluetooth kitchen thermometers.
Label printers¶
The tested Bluetooth label printer is the SUPVAN T50M family (including the T50M Pro), set up right from Settings, Printing. Plain USB and network printers that support driverless (IPP Everywhere) printing, and Zebra ZPL label printers, also work through the standard print system. See Label printing.
AI providers and hardware¶
AI features are optional. If you want fully local AI (Ollama), you need significantly more RAM (16 GB+ recommended) and ideally x86-64 with a capable CPU/GPU; vision models are slow on SBCs. Otherwise, configure a cloud provider (Gemini/OpenAI/Anthropic) in the setup wizard and any supported board above is fine.