MAI/docs/superpowers/plans/2026-06-29-docker-offline-gpu-proto-udp-implementation.md

Docker Offline GPU + Proto UDP Implementation Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: Добавить офлайн GPU Docker-упаковку для live/video запуска и новый опциональный UDP backend испытательного протокола.

Architecture: Конфигурация контейнера будет управляться env-переменными поверх текущих дефолтов в config.py и config_intercept.py. Новый backend proto_udp будет встроен в существующий AutopilotBridge, а Docker-артефакты обеспечат офлайн runtime с уже установленными зависимостями и моделью.

Tech Stack: Python 3.12, PyTorch CUDA, Ultralytics YOLO, OpenCV, Docker, docker-compose, UDP sockets.

---

Task 1: Add tests for protocol backend and env overrides

Files:

• Create: tests/test_autopilot_proto_udp.py

• Create: tests/test_runtime_env.py

• Step 1: Write failing tests for UDP packet packing and backend selection

• Step 2: Run targeted pytest commands and confirm failure

• Step 3: Add failing tests for env override parsing

• Step 4: Run targeted pytest commands and confirm failure

Task 2: Implement runtime env overrides

Files:

• Create: runtime_env.py

• Modify: config.py

• Modify: config_intercept.py

• Step 1: Add minimal env parsing helpers and runtime override API

• Step 2: Apply overrides to source/model/output/backend-related settings

• Step 3: Re-run env override tests until green

Task 3: Implement protocol UDP backend

Files:

• Modify: autopilot_bridge.py

• Modify: config_intercept.py

• Step 1: Add protocol config defaults

• Step 2: Add packet builder and proto_udp backend

• Step 3: Wire backend selection into AutopilotBridge

• Step 4: Re-run protocol tests until green

Task 4: Make runtime self-contained for container use

Files:

• Create: bytetrack_min_aggressive.py

• Modify: main.py

• Step 1: Vendor the local ByteTrack implementation into the project root

• Step 2: Remove dependency on parent-directory import for container runtime

• Step 3: Re-run affected tests and compile checks

Task 5: Add Docker packaging

Files:

• Create: Dockerfile

• Create: .dockerignore

• Create: requirements-docker.txt

• Create: docker-compose.yml

• Create: docker/entrypoint.sh

• Create: README_docker.md

• Step 1: Define offline-friendly GPU image build

• Step 2: Add compose service with GPU/runtime env setup

• Step 3: Document build and launch modes for camera/rtsp/udp/video

Task 6: Verify end-to-end artifacts

Files:

• Modify: docs/superpowers/plans/2026-06-29-docker-offline-gpu-proto-udp-implementation.md

• Step 1: Run targeted pytest for new and touched tests

• Step 2: Run python -m compileall for changed Python files

• Step 3: Report any remaining runtime limits that require host-side NVIDIA/Docker setup