Agent-Driven Development Showcases

dx-agent-dev Showcases

Real apps built on the DEEPX NPU SDK by an AI coding agent from a single natural-language prompt — each checked in with the prompt, measured results, a one-command reproduce, and the full recorded build-session transcript.

These demonstrate dx-agent-dev (Beta): you describe the app/task in plain language and the agent drives the DEEPX knowledge base end to end (brainstorm → plan → TDD → verify). What the feature is and how it works → Agent-Driven Development docs. Each card below links to that showcase's own README (full detail + transcript).

Note on build time & cost — The build time, output-token count, and cost shown for each showcase are actual measurements taken from that showcase's recorded build-session transcript (*-session.md), priced with the coding agent's published billing policy — not estimates, and not cherry-picked best runs. Because AI coding agents are non-deterministic, the same prompt does not always consume the same output tokens, so your own build time and cost may differ depending on environment and model version.

NPU-powered AI apps (mini-games)

Build a fully autonomous DEEPX-NPU app from natural language — in ~20 minutes, for ~$10. Pose-driven mini-games with arcade HUDs, built end to end from a single prompt.

Showcase	Kind	Highlight
Squat-Counting Mini-Game	game	pose game + arcade HUD
Stretching Coach Mini-Game	game	coach avatar + 3 stages

Squat-Counting Mini-Game

Counts squat reps from knee/hip angles with an arcade HUD (reps / score / DOWN·UP·GOOD!).

Highlight: pose game + arcade HUD · Claude Opus 4.8 · ≈ 12 min · ≈ $7.3 — details →

Stretching Coach Mini-Game

Guides 3 stretches with an animated coach avatar that demonstrates each target pose.

Highlight: coach avatar + 3 stages · Claude Opus 4.8 · ≈ 15 min · ≈ $8.1 — details →

Ultralytics ecosystem integration

Take any Ultralytics YOLO to the DEEPX NPU in one command — or retrain it for your domain — all in natural language. format=deepx export + 4-way eval (base/retrained × fp32-GPU / INT8-NPU); INT8 ≈ fp32, and the domain model runs faster on the NPU.

Showcase	Kind	Highlight
Ultralytics YOLO → DeepX Export	export	1-cmd .pt → .dxnn
African Wildlife Monitoring	retrain	mAP ~0.0007→0.79, 59→80 FPS
Construction PPE Safety	retrain	mAP 0.0001→0.257, 58→80 FPS
Brain-Tumor Screening	retrain	mAP ~0.0005→0.40, 59→83 FPS
Pharmaceutical Pill Inspection	retrain	mAP ~0.001→0.75 (mAP50 0.97), 55→78 FPS

Ultralytics YOLO → DeepX Export

Turns an Ultralytics YOLO .pt into a deployable DeepX NPU model (.dxnn) in a single yolo export ... format=deepx command, then runs NPU inference + verify.

Highlight: 1-cmd .pt → .dxnn · Claude Sonnet 4.6 · ≈ 12 min · ≈ $2.4 — details →

African Wildlife Monitoring

Retrains yolo26n on african-wildlife (buffalo/elephant/rhino/zebra) for a safari/conservation camera; 4-way eval base/retrained × fp32/INT8.

Highlight: mAP ~0.0007→0.79, 59→80 FPS · Claude Opus 4.8 · ≈ 7 min · ≈ $3.2 — details →

Construction PPE Safety

Retrains yolo26n on construction-ppe for a site-safety camera (helmet/vest/...); 4-way eval base/retrained × fp32/INT8.

Highlight: mAP 0.0001→0.257, 58→80 FPS · Claude Opus 4.8 · ≈ 17 min · ≈ $4.0 — details →

Brain-Tumor Screening

Retrains yolo26n on brain-tumor (MRI/CT) for a medical edge device; 4-way eval base/retrained × fp32/INT8.

Highlight: mAP ~0.0005→0.40, 59→83 FPS · Claude Opus 4.8 · ≈ 9 min · ≈ $3.7 — details →

Pharmaceutical Pill Inspection

Retrains yolo26n on medical-pills for a pharma counting station; 4-way eval base/retrained × fp32/INT8.

Highlight: mAP ~0.001→0.75 (mAP50 0.97), 55→78 FPS · Claude Opus 4.8 · ≈ 8 min · ≈ $5.1 — details →

PaddlePaddle ecosystem integration

PaddleOCR (PP-OCRv5) on the DEEPX NPU — real-time video & webcam OCR from a single, concise prompt. Baidu's PaddlePaddle OCR (text detection → orientation → recognition) running on the DX-M1 NPU.

Showcase	Kind	Highlight
Video / Webcam OCR (PP-OCRv5)	app	PP-OCRv5 det→cls→rec on-device (~2.8 FPS, 341 ms/frame); 14 text regions/frame; video + webcam from one --source flag

Video / Webcam OCR (PP-OCRv5)

Real-time text detection + recognition on the DX-M1 NPU — one code path for a video file and a live webcam, overlaying detected boxes + recognized strings.

Highlight: PP-OCRv5 det→cls→rec on-device (~2.8 FPS, 341 ms/frame); 14 text regions/frame; video + webcam from one --source flag · Claude Opus 4.8 · ≈ 18 min · ≈ $12.0 — details →

RapidAI ecosystem integration

A PDF → Markdown document-conversion app on the DEEPX NPU — from a single, concise natural-language prompt. RapidAI's RapidDoc (PP-StructureV3): layout, OCR, tables, formulas — running PaddlePaddle-trained models on the DX-M1 NPU. A standalone, self-contained app generated from the fork's pipeline.

Showcase	Kind	Highlight
PDF → Markdown (document conversion app)	app	9-page financial report parsed on-device; standalone app — vendored rapid_doc, own entry, no fork clone; 21 headings + 9 HTML tables preserved (auto 12.6s / ocr 14.7s)

PDF → Markdown (document conversion app)

Converts a PDF (digital or scanned) to structured Markdown + JSON — layout analysis, OCR, tables and formulas — on the DEEPX DX-M1 NPU via the RapidDoc fork. Supports --parse-method auto|txt|ocr.

Highlight: 9-page financial report parsed on-device; standalone app — vendored rapid_doc, own entry, no fork clone; 21 headings + 9 HTML tables preserved (auto 12.6s / ocr 14.7s) · Claude Opus 4.8 · ≈ 12 min · ≈ $6.2 — details →

Reproduce any showcase

cd dx-agent-dev-showcase/<showcase>
bash setup.sh && bash run.sh        # retrain/export showcases
## games: ./setup.sh then ./run.sh (or ./run.sh --camera 0)

Requires x86-64 Linux + the DeepX runtime (dx_engine). Per-showcase prerequisites and the exact prompt are in each showcase's README.

Korean: README-ko.md.