Feat: proof of views compute and relay to server complete.

.github/labeler.yml

@@ -4,7 +4,7 @@
   tap:
     - "tap-module/**"
   relay:
-    - "relay-node/**"
+    - "compute-node/**"
   docs:
     - "README.md"
     - "LICENSE"

.github/workflows/labeler.yml

@@ -4,6 +4,10 @@ on:
   pull_request:
     types: [opened, synchronize, reopened]
 
+permissions:
+  contents: read
+  pull-requests: write
+
 jobs:
   label:
     runs-on: ubuntu-latest
@@ -14,3 +18,4 @@ jobs:
         uses: actions/labeler@v4
         with:
           repo-token: ${{ secrets.GITHUB_TOKEN }}
+          configuration-path: .github/labeler.yml

compute-node/README.md

@@ -0,0 +1,80 @@
+# AdNet Protocol - Smart Advertising with IoT + AI
+
+**Flow**: ESP32-CAM → Stream Video → Detection using RNN and duplication avoidance using IoU → Aggregated # of viewers in 5 min durations → Send to ThingSpeak.
+
+---
+
+## 📸 ESP32-CAM Module: AdNetCamModule
+
+A tiny AI Thinker ESP32-CAM module powers our eyes-on-the-ground (literally). It streams MJPEG video over Wi-Fi so we can detect who's watching those juicy ads in real-time.
+
+### 🔧 Hardware Used
+- **Board:** ESP32-CAM (AI Thinker variant with PSRAM)
+- **Camera:** OV2640
+- **Status LED Pin:** GPIO 33
+
+### ⚙️ Features
+- MJPEG video streaming over Wi-Fi (on http://192.168.1.43/)
+- Static IP config for consistency
+- Status LED blinks while connecting, stays off when ready
+- Stream served as `multipart/x-mixed-replace;boundary=123456789000000000000987654321`
+- Optimized for high-quality JPEG (UXGA or SVGA)
+
+### 📝 Code Summary
+See `AdNetCamModule.ino` for full source. Highlights:
+- Initializes camera with dynamic config (PSRAM-aware)
+- Sets up MJPEG stream endpoint `/`
+- Streams images using ESP-IDF HTTP server
+- Loop does nothing; all work is done in background
+
+**To Access the Stream:**
+```bash
+http://192.168.1.43/  # Served MJPEG Stream
+```
+
+---
+
+## 🤖 Backend: Go + OpenCV + AI = ✨
+
+### 🔄 Flow
+1. Go app connects to MJPEG stream from ESP32-CAM
+2. Every 5 seconds, it grabs a frame
+3. Runs OpenCV DNN face detection (SSD + ResNet)
+4. Tracks new viewers (avoids counting duplicates using IoU filtering)
+5. Aggregates counts for 5 minutes
+6. Sends the count to [ThingSpeak](https://thingspeak.com/) via API
+
+### 🧩 Features
+- Uses GoCV (Go bindings for OpenCV)
+- DNN-based detector (ResNet SSD) for better accuracy than Haar
+- Deduplicates people using bounding-box IoU tracking
+- Configuration via `.env` (stream URL, model paths, thresholds)
+- Displays debug video with rectangles and confidence
+- Uses ThingSpeak to report proof-of-view metrics
+
+### 🛠⃣ How It Works
+- On every frame, we detect faces
+- We compare new bounding boxes to the previous frame
+- If IoU < threshold, it’s a new person (so we count it!)
+- Every 5 minutes, we send the total to ThingSpeak
+
+### 👀 Want to See the AI in Action?
+You’ll get a debug window with live rectangles drawn over faces along with confidence scores.
+
+---
+
+## 📍 The Result?
+- Transparent, real-world Proof of Views
+- Performance-based rewards for Ad Service Providers
+- Smarter ads that actually engage people
+- All running on open tech you control
+
+---
+
+## ✅ What's Next?
+- Plug into smart contracts for dynamic ad payouts
+- Integrate cryptography for verifiability on-chain
+
+Stay tuned. This is just the beginning.
+
+Made with ❤️ by Anon.