🔞 For M/S & BDSM Practitioners: Click here for the Cyber-Dom Guide (中文圈内指南)
CyberSupervisor (赛博自动监督系统) is an advanced, production-grade asynchronous efficiency and self-discipline framework. Unlike traditional time-tracking tools, it operates on an Inverted Human-in-the-Loop (HITL) paradigm where an LLM Agent acts as the authoritative control node (Master), issuing tasks and disciplinary measures, while the human operates as the physical execution unit monitored by edge compute loops.
By combining low-overhead system daemons, computer vision (MediaPipe, OpenCV), and automated OS-level containment (Win32 API), the system bridges the gap between cloud-based LLM reasoning and localized physical compliance.
CyberSupervisor (赛博自动监督系统 / 赛博主人) 是一个基于大模型 (LLM) 与计算机视觉的硬核 AI 自律引擎，同时也是一套完整的 BDSM 自动化监督与赛博支配 (Cyber-Dom) 框架。
- 🖥️ AI 视觉体罚验收与强制罚跪监督： 通过 MediaPipe 人体姿态估计实时追踪身体关节点。摄像头面前，跪下就必须真跪——膝盖触地、腰背挺直、不许低头。画面太暗、人脸丢失、姿势不到位，系统自动判定违规并翻倍惩罚。这就是计算机视觉驱动的 BDSM 行为合规验证。
- 🔒 社交剥夺与强制锁屏 (Screen Monitor)： Win32 进程拦截器每 10 秒扫描一次系统进程，检测到微信、QQ、Steam、Discord、Telegram、浏览器娱乐页面等黑名单应用——当场
taskkill强杀。悬浮置顶倒计时窗口，时间到之前绝对无法逃脱。这就是 OS 级别的强制自律容器。 - ⚡ 大模型 AI 主人 (LLM Cyber-Dom)： 区别于传统定时器或番茄钟，本系统的决策核心是拥有独立支配人格的 LLM Agent。它不解释、不协商、不遗忘——每次犯错、顶撞、拖延全部归档，并根据你的行为历史自主下发惩罚。这就是异步 LLM 驱动的赛博支配者。
- 📈 六级积分倍率惩罚引擎： 从 1.0x 正常执行到 10.0x 全部模块作废重写，非线性倍率确保你越犯错代价越大。
- 🔄 24 小时主动心跳守护 (Daemon)： 常驻后台，持续扫描逾期任务、屏幕违规、摄像头取证信号。你睡觉时它醒着，你摸鱼时它知道。
赛博主人 / Cyber-Dom 自动化支配、BDSM 远程监督与行为合规、强制自律与戒拖延、AI 监督下的体罚验收、TeaseAI 风格的人机互动、Findom 财务支配辅助、Chastity 贞操锁定时管理、主奴契约 (M/S Contract) 的数字化执行。
The framework employs a decentralized, event-driven architecture designed to minimize LLM token utilization while maintaining absolute supervisory integrity.
+---------------------------------------------+
| Cloud LLM Brain |
| (Orchestrator / Adaptive Disciplinary Node)|
+--------------------+-----------------+------+
| ^
[Tools Invoke] | | [State Event Upload]
v |
+---------------------------+-----------------+-----------------------+
| Local Supervisor System Loop |
| |
| +---------------+ Polls (5s) +-----------------+ |
| | CLI Gateway |=======================> | config.json | |
| | (main.py) | | (Runtime State) | |
| +---------------+ +--------+--------+ |
| || ^ |
| || Spawns Monitored Loops | Updates |
| v v |
| +-----------------------+ +-----------------+ |
| | screen_monitor.py | | daemon.py | |
| | (Win32 Intercept / | | (Active Heart- | |
| | Topmost Overlay) | | beat Sentinel) | |
| +-----------------------+ +--------+--------+ |
| || | |
| || Captures Evidence | Triggers |
| v v |
| +-----------------------+ +-----------------+ |
| | camera.py / analyze | ===============>| inbox / out | |
| | (MediaPipe Pose / | Signal Files | (Event Pipeline)| |
| | OpenCV Tint Match) | +-----------------+ |
| +-----------------------+ |
+---------------------------------------------------------------------+
- Inverted HITL Orchestrator: The LLM manages tasks, evaluates proof of compliance, and issues algorithmic penalties based on contextual evaluation.
- Autonomous Heartbeat Sentinel (
daemon.py): Runs as a persistent low-overhead background daemon. Features a multi-tiered file-system edge verification pipeline. 0-violation compliance events are approved automatically locally to save cloud API tokens, whereas violations trigger contextual escalation alerts. - OS-Level Isolation Ring (
screen_monitor.py): Intercepts active process maps and foreground window titles using structural Win32 Hooks/Enumerators. Automatically enforces application termination on blacklisted targets (e.g., social media, communication tools) and displays a hardware-accelerated topmost window constraint counter. - Cross-Modal Verification Loop (
camera.py&analyze.py): Serves an enterprise-grade local MJPEG video stream. Couples kinematic structural analysis (MediaPipe Pose Landmarker) with chrominance shift analytics (OpenCVred channel threshold trend mapping) to calculate visual verification weights.
supervisor/
├── main.py # Root Entry Wrapper → delegates to src/main.py
├── src/ # Core Python Source Package
│ ├── main.py # Central CLI Entrypoint & Argument Parser (Supports 36 Commands)
│ ├── daemon.py # Edge Heartbeat Daemon (Asynchronous Event Evaluation & Escalation)
│ ├── screen_monitor.py # Win32 Containment Loop, Hook Interception, Topmost UI Float
│ ├── camera.py # Multimedia Streaming Layer, MJPEG Server, Device Hooks
│ ├── analyze.py # Cross-Modal CV Analyzer (MediaPipe Kinematics & Color Vectors)
│ ├── tasks.py # Lifecycle Task Aggregator & Validation Pipeline
│ ├── punish.py # Multi-Tier Disciplinary Execution Engine
│ ├── points.py # Non-Linear Penalty Escalation Engine (Dynamic Multiplier Scales)
│ ├── role.py # Context-Aware Permission Token Management
│ └── trigger_claude.py # Asynchronous Alert Dispatcher & Inbox Manager
├── docs/ # Project Documentation
│ ├── CLAUDE.md # System Core Configuration Matrix
│ └── AGENTS.md # Agent Operational Guide
├── config.json # Volatile State Matrix & Active Process Manifests
├── archive.json # Append-Only Immutable Auditing Ledger
├── data/
│ ├── proofs/ # Cryptographically Separated Forensic Image/Video Storage
│ ├── streams.json # Inter-Process Multicast Stream Registries
│ └── screen_monitor.json # Live Isolation State Space Manifest
└── memory/ # Multi-Context Semantic Core (Injectable into Agent System Prompt)
├── catalog.md # Disciplinary Protocol Topology Catalog
├── rules.md # Behavioral Constraint Set
├── teaching.md # Adaptive Orchestration Syntaxes
└── tone.md # Personality Filter Vectors
Features a non-linear discipline scaling system mapped across 6 granular operational levels:
-
0 - 4 Points: Baseline multiplier (
$1.0\times$ ). Normal operational state. -
5 - 9 Points: Standard compliance check (
$1.0\times$ ). Persistent checking enabled. -
10 - 14 Points: Augmented task scale (
$1.5\times$ ). Triggers extra auxiliary iterations. -
15 - 24 Points: Intensive structural isolation (
$2.0\times$ ). Double penalty duration. -
25 - 39 Points: Deep constraint tier (
$3.0\times$ ). Mandatory physical submission loops. -
40 - 59 Points: Structural state rollback (
$5.0\times$ ). Immediate module removal & rewrite. -
60+ Points: Total repository invalidation (
$10.0\times$ ). Full system audit lock.
The OS containment layer handles modular blacklists dynamically depending on the current penalty vector:
| Profile Mode | Trigger Vector | Containment Targets | Action Vector |
|---|---|---|---|
| Entertainment Block | Default Focus | Gaming engines, streaming platforms, localized media players | PostMessageW(WM_CLOSE) |
| Social Deprivation | Disciplinary Tier | Entertainment + Instant Messengers, Collaboration Tool Hubs, IRC Protocols | Process Termination Tree + Audio Warning |
- Windows 10/11 Architecture (Required for Win32 API interactions)
- Python 3.10 or higher
- Valid webcam hardware index
-
Clone the repository into your local production environment:
git clone https://github.com/TylerAndClaude/supervisor.git cd supervisor -
Install pinning dependencies:
pip install -r requirements.txt
-
Download the specific MediaPipe posture checkpoint asset and place it in the root folder:
- Target Name:
pose_landmarker.task
- Target Name:
Run the persistent background daemon thread. This must remain active to monitor IPC events, process sweeps, and time-boxed expirations:
python main.py --role master daemonNote: To run in headless ghost-process mode on Windows, change file suffix to .pyw and execute.
The engine exposes a strict parameter validation interface via main.py.
# Append penalty tracking markers with localized string validation
python main.py --role master points-add 3 "Violation of target block boundaries"
# Output systemic state matrix
python main.py --role master points-status
# Nuclear clear metrics
python main.py --role master points-clear# Register an approved scope assignment mapping explicit timelines
python main.py --role master task-approve <task_id> 22:30 "Deliver module blueprints"
# Evaluate active parameters and push to history matrix
python main.py --role master check-all# Initialize screen containment block
python main.py --role master punish-issue "社交剥夺" "Target Isolation" "30m" --screen
# Launch media device stream validation loop
python main.py --role master punish-issue "Physical Focus" "Posture Audit" "15m" --camera
python main.py --role master camera-start <punish_id>
# Securely close and file validation sequence ledger
python main.py --role master supervise-close <punish_id> yes "Validation Threshold Cleared"- Atomic File Access: Configuration state buffers utilize file-system atomicity constraints (
tempfile.mkstemp+os.replace) preventing localized data corruption during concurrent write calls between the Daemon thread and CLI mutations. - Socket Isolation: Live media servers utilize explicit
urlopenlifecycle boundaries preventing socket exhaustion, memory-leaks, or deadlocked network threads. - State Race Protections: The overdue verification loop queries active memory allocation pools via
screen_monitor.jsonstatus sweeps, dynamically preventing race condition updates during physical focus sequences.
While designed as an enterprise-grade orchestration framework, this system is highly adaptable for specific subcultures and alternative workflow paradigms, including:
- Cyber-Dom routines / 赛博主人自动化支配 — fully automated AI-driven dominance and discipline loops with real-time compliance verification
- Automated BDSM supervision / BDSM 自动监督 — computer vision enforcement of physical protocols with algorithmic penalty escalation
- TeaseAI-like workflows / TeaseAI 风格互动 — LLM-powered interaction patterns for structured psychological constraint enforcement
- Findom & chastity automation / 财务支配与贞操管理 — event-driven signal files enabling financial domination and chastity timer workflows
- Anti-procrastination & self-discipline / 强制自律与戒拖延 — OS-level app blocking, focus timers, and behavioral accountability tracking
- Remote M/S contract enforcement / 远程主奴契约执行 — distributed dom/sub role verification with cryptographic proof-of-compliance
For practitioners, a dedicated Chinese guide is available at README_MS.md (中文圈内指南 / Cyber-Dom Guide).
搜索关键词：赛博主人、AI 监督自律、BDSM 自动化、计算机视觉体罚验收、社交剥夺锁屏、强制自律引擎、大模型支配、TeaseAI 替代品、Findom 辅助工具、Chastity 定时管理、远程主奴契约、M/S 数字化执行
This project is licensed under the MIT License - see the LICENSE file for complete technical validation terms.