Skip to content

Commit 91c6c42

Browse files
solderzzcsolderzzc
committed
feat: migrate Coral TPU skill from pycoral to ai-edge-litert (LiteRT)
Breaking change: Docker is no longer the default deployment — native venv is. detect.py: - Replace pycoral + tflite-runtime with ai-edge-litert - Use litert.Interpreter + load_delegate('libedgetpu') for Edge TPU - Platform-aware delegate lib name (Linux/macOS/Windows) - Unified output tensor parsing (removed separate pycoral path) - Add 'runtime' field to ready event requirements.txt: - pycoral~=2.0 + tflite-runtime~=2.14 → ai-edge-litert>=2.1.0 - Python 3.9-3.13 supported (was 3.6-3.9) deploy.sh: - Docker build → native Python venv - Auto-installs libedgetpu via apt on Debian/Ubuntu - Install instructions for macOS and Windows - Native run_command (no docker run) deploy.bat: - Docker → native Python venv for Windows tpu_probe.py: - list_edge_tpus() → load_delegate() availability test SKILL.md: - Version 1.0.0 → 2.0.0 - Removed runtime: docker - Updated Platform Setup with libedgetpu install commands - Docker noted as optional alternative Dockerfile: - Marked as optional, bullseye-slim → bookworm-slim - pycoral → ai-edge-litert
1 parent c8ae2ac commit 91c6c42

7 files changed

Lines changed: 387 additions & 335 deletions

File tree

‎skills/detection/yolo-detection-2026-coral-tpu/Dockerfile‎

Lines changed: 10 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -1,13 +1,15 @@
1-
# ─── Coral TPU Detection — Runtime Image ─────────────────────────────────────
2-
# Runs YOLO inference on Google Coral Edge TPU via pycoral/tflite-runtime.
3-
# Built locally on user's machine. No external model registry.
1+
# ─── Coral TPU Detection — Docker Image (Optional) ──────────────────────────
2+
# Alternative to native deployment for environments where installing
3+
# libedgetpu or managing Python venvs is impractical.
4+
#
5+
# PREFERRED: Use deploy.sh for native installation (no Docker needed).
46
#
57
# Build: docker build -t aegis-coral-tpu .
68
# Run: docker run -i --rm --device /dev/bus/usb \
79
# -v /tmp/aegis_detection:/tmp/aegis_detection \
810
# aegis-coral-tpu
911

10-
FROM debian:bullseye-slim
12+
FROM debian:bookworm-slim
1113

1214
# Avoid interactive prompts
1315
ENV DEBIAN_FRONTEND=noninteractive
@@ -17,15 +19,15 @@ ENV PYTHONUNBUFFERED=1
1719
RUN apt-get update && apt-get install -y --no-install-recommends \
1820
python3 \
1921
python3-pip \
20-
python3-dev \
22+
python3-venv \
2123
gnupg2 \
2224
ca-certificates \
2325
curl \
2426
usbutils \
2527
libusb-1.0-0 \
2628
&& rm -rf /var/lib/apt/lists/*
2729

28-
# ─── Add Google Coral package repository ─────────────────────────────────────
30+
# ─── Add Google Coral package repository (libedgetpu) ────────────────────────
2931
RUN echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" \
3032
> /etc/apt/sources.list.d/coral-edgetpu.list \
3133
&& curl -fsSL https://packages.cloud.google.com/apt/doc/apt-key.gpg \
@@ -35,9 +37,9 @@ RUN echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" \
3537
libedgetpu1-std \
3638
&& rm -rf /var/lib/apt/lists/*
3739

38-
# ─── Python dependencies ─────────────────────────────────────────────────────
40+
# ─── Python dependencies (ai-edge-litert replaces pycoral) ──────────────────
3941
COPY requirements.txt /app/requirements.txt
40-
RUN pip3 install --no-cache-dir -r /app/requirements.txt
42+
RUN pip3 install --no-cache-dir --break-system-packages -r /app/requirements.txt
4143

4244
# ─── Application code ────────────────────────────────────────────────────────
4345
COPY scripts/ /app/scripts/

‎skills/detection/yolo-detection-2026-coral-tpu/SKILL.md‎

Lines changed: 39 additions & 38 deletions
Original file line numberDiff line numberDiff line change
@@ -1,14 +1,15 @@
11
---
22
name: yolo-detection-2026-coral-tpu
3-
description: "Google Coral Edge TPU — real-time object detection via Docker"
4-
version: 1.0.0
3+
description: "Google Coral Edge TPU — real-time object detection with LiteRT"
4+
version: 2.0.0
55
icon: assets/icon.png
66
entry: scripts/detect.py
77
deploy: deploy.sh
8-
runtime: docker
98

109
requirements:
11-
docker: ">=20.10"
10+
python: ">=3.9"
11+
ai-edge-litert: ">=2.1.0"
12+
system: "libedgetpu"
1213
platforms: ["linux", "macos", "windows"]
1314

1415
parameters:
@@ -78,59 +79,56 @@ mutex: detection
7879

7980
# Coral TPU Object Detection
8081

81-
Real-time object detection using Google Coral Edge TPU accelerator. Runs inside Docker for cross-platform support. Detects 80 COCO classes (person, car, dog, cat, etc.) with ~4ms inference on 320×320 input.
82+
Real-time object detection using Google Coral Edge TPU accelerator. Uses [ai-edge-litert](https://pypi.org/project/ai-edge-litert/) (LiteRT — the modern successor to TFLite/pycoral) with the `libedgetpu` delegate for hardware acceleration. Detects 80 COCO classes with ~4ms inference on 320×320 input.
8283

8384
## Requirements
8485

8586
- **Google Coral USB Accelerator** (USB 3.0 port recommended)
86-
- **Docker Desktop 4.35+** (all platforms — Linux, macOS, Windows)
87-
- USB 3.0 cable (the included cable is recommended)
87+
- **Python 3.9+** (3.9–3.13 supported via ai-edge-litert)
88+
- **libedgetpu** system library (installed per platform, see below)
8889
- Adequate cooling for sustained inference
8990

9091
## How It Works
9192

92-
```
93-
┌─────────────────────────────────────────────────────┐
94-
│ Host (Aegis-AI) │
95-
│ frame.jpg → /tmp/aegis_detection/ │
96-
│ stdin ──→ ┌──────────────────────────────┐ │
97-
│ │ Docker Container │ │
98-
│ │ detect.py │ │
99-
│ │ ├─ loads _edgetpu.tflite │ │
100-
│ │ ├─ reads frame from volume │ │
101-
│ │ └─ runs inference on TPU │ │
102-
│ stdout ←── │ → JSONL detections │ │
103-
│ └──────────────────────────────┘ │
104-
│ USB ──→ /dev/bus/usb (Linux) or USB/IP (Mac/Win) │
105-
└─────────────────────────────────────────────────────┘
106-
```
107-
108-
1. Aegis writes camera frame JPEG to shared `/tmp/aegis_detection/` volume
109-
2. Sends `frame` event via stdin JSONL to Docker container
110-
3. `detect.py` reads frame, runs inference on Edge TPU
111-
4. Returns `detections` event via stdout JSONL
112-
5. Same protocol as `yolo-detection-2026` — Aegis sees no difference
93+
1. `deploy.sh` creates a Python venv, installs `ai-edge-litert`, and checks for `libedgetpu`
94+
2. Aegis writes camera frame JPEG to shared `/tmp/aegis_detection/` directory
95+
3. Sends `frame` event via stdin JSONL to `detect.py`
96+
4. `detect.py` loads the Edge TPU delegate via `litert.load_delegate('libedgetpu')`
97+
5. Returns `detections` event via stdout JSONL
98+
6. Same protocol as `yolo-detection-2026` — Aegis sees no difference
11399

114100
## Platform Setup
115101

116102
### Linux
117103
```bash
118-
# USB Coral should be auto-detected
119-
# Docker uses --device /dev/bus/usb for direct access
104+
# 1. Install libedgetpu (deploy.sh auto-installs on Debian/Ubuntu)
105+
echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" | \
106+
sudo tee /etc/apt/sources.list.d/coral-edgetpu.list
107+
curl -fsSL https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -
108+
sudo apt-get update && sudo apt-get install libedgetpu1-std
109+
110+
# 2. Connect Coral USB Accelerator, then:
120111
./deploy.sh
121112
```
122113

123-
### macOS (Docker Desktop 4.35+)
114+
### macOS
124115
```bash
125-
# Docker Desktop USB/IP handles USB passthrough
126-
# ARM64 Docker image builds natively on Apple Silicon
116+
# 1. Install libedgetpu
117+
curl -LO https://github.com/google-coral/libedgetpu/releases/download/release-grouper/edgetpu_runtime_20221024.zip
118+
unzip edgetpu_runtime_20221024.zip && cd edgetpu_runtime && sudo bash install.sh
119+
120+
# 2. Connect Coral USB Accelerator, then:
127121
./deploy.sh
128122
```
129123

130124
### Windows
131125
```powershell
132-
# Docker Desktop 4.35+ with USB/IP support
133-
# Or WSL2 backend with usbipd-win
126+
# 1. Install libedgetpu
127+
# Download edgetpu_runtime_20221024.zip from:
128+
# https://github.com/google-coral/libedgetpu/releases
129+
# Extract and run install.bat
130+
131+
# 2. Connect Coral USB Accelerator, then:
134132
.\deploy.bat
135133
```
136134

@@ -139,7 +137,6 @@ Real-time object detection using Google Coral Edge TPU accelerator. Runs inside
139137
Ships with pre-compiled `yolo26n_edgetpu.tflite` (YOLO 2026 nano, INT8 quantized, 320×320). To compile custom models:
140138

141139
```bash
142-
# Requires x86_64 Linux or Docker --platform linux/amd64
143140
python scripts/compile_model.py --model yolo26s --size 320
144141
```
145142

@@ -158,7 +155,7 @@ Same JSONL as `yolo-detection-2026`:
158155

159156
### Skill → Aegis (stdout)
160157
```jsonl
161-
{"event": "ready", "model": "yolo26n_edgetpu", "device": "coral", "format": "edgetpu_tflite", "tpu_count": 1, "classes": 80}
158+
{"event": "ready", "model": "yolo26n_edgetpu", "device": "coral", "format": "edgetpu_tflite", "runtime": "ai-edge-litert", "tpu_count": 1, "classes": 80}
162159
{"event": "detections", "frame_id": 42, "camera_id": "front_door", "objects": [{"class": "person", "confidence": 0.85, "bbox": [100, 50, 300, 400]}]}
163160
{"event": "perf_stats", "total_frames": 50, "timings_ms": {"inference": {"avg": 4.1, "p50": 3.9, "p95": 5.2}}}
164161
```
@@ -172,4 +169,8 @@ Same JSONL as `yolo-detection-2026`:
172169
./deploy.sh
173170
```
174171

175-
The deployer builds the Docker image locally, probes for TPU devices, and sets the runtime command. No packages pulled from external registries beyond Docker base images and Coral apt repo.
172+
The deployer creates a Python venv, installs `ai-edge-litert` and dependencies, checks for the `libedgetpu` system library, probes for Edge TPU devices, and sets the native `run_command`.
173+
174+
### Docker (Optional)
175+
176+
A `Dockerfile` is provided for environments where native installation is impractical. Note: Docker requires USB/IP passthrough for Edge TPU access on macOS and Windows.

‎skills/detection/yolo-detection-2026-coral-tpu/deploy.bat‎

Lines changed: 71 additions & 46 deletions
Original file line numberDiff line numberDiff line change
@@ -1,71 +1,96 @@
11
@echo off
2-
REM deploy.bat — Docker-based bootstrapper for Coral TPU Detection Skill (Windows)
2+
REM deploy.bat — Native bootstrapper for Coral TPU Detection Skill (Windows)
33
REM
4-
REM Builds the Docker image locally and verifies Edge TPU connectivity.
4+
REM Installs ai-edge-litert in a Python venv and verifies Edge TPU connectivity.
55
REM Called by Aegis skill-runtime-manager during installation.
66
REM
7-
REM Requires: Docker Desktop 4.35+ with USB/IP support
7+
REM Prerequisites:
8+
REM - Python 3.9+ installed and on PATH
9+
REM - libedgetpu installed (https://github.com/google-coral/libedgetpu/releases)
10+
REM
11+
REM Exit codes:
12+
REM 0 = success (Edge TPU detected)
13+
REM 1 = fatal error
14+
REM 2 = partial success (no TPU, CPU fallback)
15+
16+
setlocal EnableDelayedExpansion
817

9-
setlocal enabledelayedexpansion
18+
set SKILL_DIR=%~dp0
19+
set VENV_DIR=%SKILL_DIR%.venv
20+
set LOG_PREFIX=[coral-tpu-deploy]
1021

11-
set "SKILL_DIR=%~dp0"
12-
set "IMAGE_NAME=aegis-coral-tpu"
13-
set "IMAGE_TAG=latest"
14-
set "LOG_PREFIX=[coral-tpu-deploy]"
22+
REM ─── Step 1: Find Python ───────────────────────────────────────────────────
1523

16-
REM ─── Step 1: Check Docker ────────────────────────────────────────────────
24+
echo %LOG_PREFIX% Checking for Python 3.9+... >&2
1725

18-
where docker >nul 2>&1
19-
if %errorlevel% neq 0 (
20-
echo %LOG_PREFIX% ERROR: Docker not found. Install Docker Desktop 4.35+ 1>&2
21-
echo {"event": "error", "stage": "docker", "message": "Docker not found. Install Docker Desktop 4.35+"}
26+
where python >nul 2>&1
27+
if errorlevel 1 (
28+
echo {"event": "error", "stage": "python", "message": "Python not found. Install Python 3.9+"}
2229
exit /b 1
2330
)
2431

25-
docker info >nul 2>&1
26-
if %errorlevel% neq 0 (
27-
echo %LOG_PREFIX% ERROR: Docker daemon not running. Start Docker Desktop. 1>&2
28-
echo {"event": "error", "stage": "docker", "message": "Docker daemon not running"}
29-
exit /b 1
32+
for /f "tokens=2" %%v in ('python --version 2^>^&1') do set PY_VER=%%v
33+
echo %LOG_PREFIX% Found Python %PY_VER% >&2
34+
35+
REM ─── Step 2: Create virtual environment ────────────────────────────────────
36+
37+
if not exist "%VENV_DIR%\Scripts\python.exe" (
38+
echo {"event": "progress", "stage": "venv", "message": "Creating virtual environment..."}
39+
python -m venv "%VENV_DIR%"
40+
if errorlevel 1 (
41+
echo {"event": "error", "stage": "venv", "message": "Failed to create virtual environment"}
42+
exit /b 1
43+
)
3044
)
3145

32-
for /f "tokens=*" %%v in ('docker version --format "{{.Server.Version}}" 2^>nul') do set "DOCKER_VER=%%v"
33-
echo %LOG_PREFIX% Using Docker (version: %DOCKER_VER%) 1>&2
34-
echo {"event": "progress", "stage": "docker", "message": "Docker ready (%DOCKER_VER%)"}
46+
set VENV_PIP=%VENV_DIR%\Scripts\pip.exe
47+
set VENV_PYTHON=%VENV_DIR%\Scripts\python.exe
3548

36-
REM ─── Step 2: Build Docker image ──────────────────────────────────────────
49+
echo %LOG_PREFIX% Virtual environment ready >&2
3750

38-
echo %LOG_PREFIX% Building Docker image: %IMAGE_NAME%:%IMAGE_TAG% ... 1>&2
39-
echo {"event": "progress", "stage": "build", "message": "Building Docker image..."}
51+
REM ─── Step 3: Install dependencies ─────────────────────────────────────────
4052

41-
docker build -t %IMAGE_NAME%:%IMAGE_TAG% "%SKILL_DIR%"
42-
if %errorlevel% neq 0 (
43-
echo %LOG_PREFIX% ERROR: Docker build failed 1>&2
44-
echo {"event": "error", "stage": "build", "message": "Docker image build failed"}
45-
exit /b 1
46-
)
53+
echo {"event": "progress", "stage": "install", "message": "Installing ai-edge-litert and dependencies..."}
54+
"%VENV_PIP%" install --upgrade pip -q 2>&1
55+
"%VENV_PIP%" install -r "%SKILL_DIR%requirements.txt" -q 2>&1
4756

48-
echo {"event": "progress", "stage": "build", "message": "Docker image ready"}
57+
echo %LOG_PREFIX% Dependencies installed >&2
58+
echo {"event": "progress", "stage": "install", "message": "Dependencies installed"}
4959

50-
REM ─── Step 3: Probe for Edge TPU ──────────────────────────────────────────
60+
REM ─── Step 4: Check libedgetpu ──────────────────────────────────────────────
5161

52-
echo %LOG_PREFIX% Probing for Edge TPU devices... 1>&2
53-
echo {"event": "progress", "stage": "probe", "message": "Checking for Edge TPU devices..."}
62+
echo %LOG_PREFIX% Checking for libedgetpu... >&2
63+
echo {"event": "progress", "stage": "driver", "message": "Checking for Edge TPU driver..."}
5464

55-
docker run --rm --privileged %IMAGE_NAME%:%IMAGE_TAG% python3 scripts/tpu_probe.py >nul 2>&1
56-
if %errorlevel% equ 0 (
57-
echo %LOG_PREFIX% Edge TPU detected 1>&2
58-
echo {"event": "progress", "stage": "probe", "message": "Edge TPU detected"}
59-
) else (
60-
echo %LOG_PREFIX% WARNING: No Edge TPU detected - CPU fallback 1>&2
61-
echo {"event": "progress", "stage": "probe", "message": "No Edge TPU detected - CPU fallback"}
65+
where edgetpu.dll >nul 2>&1
66+
if errorlevel 1 (
67+
echo %LOG_PREFIX% WARNING: edgetpu.dll not found. >&2
68+
echo %LOG_PREFIX% Download the runtime from: >&2
69+
echo %LOG_PREFIX% https://github.com/google-coral/libedgetpu/releases/download/release-grouper/edgetpu_runtime_20221024.zip >&2
70+
echo %LOG_PREFIX% Extract and run install.bat >&2
71+
echo {"event": "progress", "stage": "driver", "message": "libedgetpu not found — install from Google releases"}
6272
)
6373

64-
REM ─── Step 4: Set run command ──────────────────────────────────────────────
74+
REM ─── Step 5: Probe for Edge TPU ───────────────────────────────────────────
75+
76+
echo {"event": "progress", "stage": "probe", "message": "Checking for Edge TPU devices..."}
6577

66-
set "RUN_CMD=docker run -i --rm --privileged -v /tmp/aegis_detection:/tmp/aegis_detection --env AEGIS_SKILL_ID --env AEGIS_SKILL_PARAMS --env PYTHONUNBUFFERED=1 %IMAGE_NAME%:%IMAGE_TAG%"
78+
set TPU_FOUND=false
79+
"%VENV_PYTHON%" "%SKILL_DIR%scripts\tpu_probe.py" > "%TEMP%\tpu_probe_result.json" 2>nul
80+
if not errorlevel 1 (
81+
set TPU_FOUND=true
82+
)
6783

68-
echo {"event": "complete", "status": "success", "run_command": "%RUN_CMD%", "message": "Coral TPU skill installed"}
84+
REM ─── Step 6: Complete ──────────────────────────────────────────────────────
6985

70-
echo %LOG_PREFIX% Done! 1>&2
71-
exit /b 0
86+
set RUN_CMD=%VENV_PYTHON% %SKILL_DIR%scripts\detect.py
87+
88+
if "%TPU_FOUND%"=="true" (
89+
echo {"event": "complete", "status": "success", "accelerator_found": true, "run_command": "%RUN_CMD%", "message": "Coral TPU skill installed — Edge TPU ready"}
90+
echo %LOG_PREFIX% Done! Edge TPU ready. >&2
91+
exit /b 0
92+
) else (
93+
echo {"event": "complete", "status": "partial", "accelerator_found": false, "run_command": "%RUN_CMD%", "message": "Coral TPU skill installed — no TPU detected (CPU fallback)"}
94+
echo %LOG_PREFIX% Done with warning: no TPU detected. >&2
95+
exit /b 2
96+
)

0 commit comments

Comments
 (0)