alwaysai-setup

alwaysAI on NVIDIA Jetson

alwaysAI is a computer vision development platform for creating and deploying ML applications on edge devices. Deploy object detection projects from a host PC to Jetson via SSH, with TensorRT-optimized models for real-time inference.

Hardware: Jetson device (Nano/Xavier NX/AGX Xavier/AGX Orin), USB webcam or MIPI CSI camera Software: JetPack 4.6 with all SDK components, host PC (Windows/Linux/Mac)

---

Execution model

Run one phase at a time. After each phase:

• If output contains [STOP] → stop immediately, consult the failure decision tree
• If output ends with [OK] → tell the user "Phase N complete" and proceed

---

Phase 1 — prerequisites check (~30 s)

On Jetson:

sudo apt-cache show nvidia-jetpack
# Confirm JetPack 4.6

ls /dev/video*
# Confirm camera is connected

[OK] when JetPack 4.6 confirmed and camera detected.

---

Phase 2 — setup host PC (~5 min)

On the development PC:

1. Download and install alwaysAI from https://alwaysai.co/installer/windows (or Mac/Linux equivalent)
2. Verify CLI:

aai -v

1. Verify OpenSSH:

ssh -V

[OK] when aai and ssh both return version numbers.

---

Phase 3 — setup Jetson environment (~2 min)

On Jetson:

sudo usermod -aG docker $USER

Log out and back in, then verify:

docker run hello-world

[OK] when hello-world runs without sudo.

---

Phase 4 — create account & project (human action)

1. Sign up at https://console.alwaysai.co/auth?register=true
2. Create a new project: Dashboard → New Project → Object Detection
3. Delete the default mobilenet_ssd model (not optimized for Jetson)
4. Add optimized model: Model Catalog → search ssd_mobilenet_v1_coco_2018_01_28_xavier_nx → Add To Project

[OK] when project has the TensorRT-optimized model.

---

Phase 5 — deploy to Jetson (~5 min)

On host PC, create a project folder and configure:

mkdir ~/alwaysai-project && cd ~/alwaysai-project
aai app configure