Demo 06 — Run the alpha without hardware

The complete operator UI on a laptop, driven by synthetic scenes. Five minutes, no PCIe, no camera.

Needs: any Linux/macOS/Windows machine with Python 3.11, the repository. No Akida, no camera.

Environment: a plain venv with numpy<2, opencv-python; pyserial optional.

Run

git clone https://github.com/<org>/vorsa-inspect && cd vorsa-inspect
python3 -m venv .venv && source .venv/bin/activate
pip install "numpy<2" opencv-python
python3 tools/run_web.py --synthetisch --port 8080

Output we got

  Anzeige:    960 px
  Daten:      vorsa_daten
  Lernstand:  noch keiner  (keine Begleitdatei)
  Hardware:   {'akida': False, 'geraet': None, 'nps': None, 'tensorflow': False,
               'cnn2snn': False, 'build': '1.9.37-lernrahmen', 'akida_fehler': "No module named 'akida'"}
  Beenden mit Strg+C

Open http://localhost:8080. The footer shows BUILD 1.9.37-lernrahmen. The image is a generated scene with moving synthetic parts; the geometric pipeline segments and frames them; the Akida pool runs in stub mode and returns a deliberately flat confidence (1/C) so that nobody mistakes it for a model.

What to click

1. Set up → Objects & learning: create an object, take photos of a synthetic part (space bar), watch the immediate photo check (sharpness, area).
2. Operation → Line: draw both zones, switch Flow, press Line start. The belt is simulated ("Simulation" in the line tile); parts enter the delivery zone, the state machine runs RUN → INSPECT → RUN → PICKUP.
3. System → Status → Hardware: "MetaTF nicht importierbar" — honest, no fake cards.
4. Try Learning zones, the camera strip, the overview zoom, the alarm book.

Proves

Server, UI, state machine, zones, recipes, alarms, inspection log and the segmentation front end work as a whole. It is the environment our UI smoke test runs against (44 checks in jsdom).

Does not prove

Anything about Akida timing, mapping, camera behaviour or recognition quality. "Chip learning" is refused without a card.

Failure modes

ModuleNotFoundError: cv2 → install opencv-python in the venv. Port in use → --port 8090. Windows: use py -3.11 and .venv\Scripts\activate.

Time: 5 minutes.