Sentinel Core

The Mission-Critical Surveillance Framework.

Sentinel Core is the open-source foundational layer of the Sentinel platform. It provides the architectural primitives required for high-performance, autonomous video intelligence, object detection, and multi-stream camera management.

This core library is maintained by AB Labs and is engineered for environments where reliability and low-latency processing are non-negotiable.

🎥 See It In Action

Watch the full demo: Natural language search across live surveillance feeds, powered by CLIP + YOLO on Apple Silicon.

---

Strategic Pipeline

graph LR
    A[Ingestion] --> B[Intelligence]
    B --> C[Spatial Awareness]
    C --> D[Alerting]
    C --> E[Persistence]
    
    style B fill:#1F4FD8,color:#fff
    style D fill:#D4AF37,color:#000

Loading

High-Level Architecture

Sentinel Core is designed as a modular pipeline:

1. Ingestion: Multi-protocol stream handling (RTSP, RTMP, Local).
2. Detection: Abstraction layers for AI model inference (YOLO, CLIP, Custom).
3. Intelligence: Data-schema for persistent tracking and situational awareness.
4. Spatial Awareness: Polygon-based zone detection, intrusion alerts, and loitering analysis.
5. Semantic Search: Natural language querying of technical data streams via sovereign CLIP embeddings.
6. Sovereign Alerting: Standardized protocols for secure alert delivery via Telegram, Webhooks, and custom channels.
7. Evidence Persistence: High-fidelity video recording with pre-event buffering and automated segment management.
8. Hardware Acceleration: Silicon-aware optimization for NVIDIA CUDA and Apple Silicon (MPS).
9. Sovereign Memory: Institutional-grade persistence for detections and alerts via local SQLite substrate.
10. Tactical Dashboard: A premium web interface for real-time monitoring and natural language forensic search.

Getting Started

Sentinel Core is designed to be highly extensible. Below is a minimal implementation of a custom situational awareness detector:

from ai.base import BaseDetector, Detection
from core.manager import StreamManager

class CustomUnitDetector(BaseDetector):
    def detect(self, frame):
        # Implementation of your AI model inference logic
        return [
            Detection(class_id=0, class_name="person", confidence=0.98, bbox=(0.1, 0.1, 0.5, 0.5))
        ]

# Initialize the Orchestrator
manager = StreamManager()
manager.set_detector(CustomUnitDetector())

# Initiate a stream session
import asyncio
asyncio.run(manager.start_stream(camera_id=101, source="rtsp://internal.secure-feed.local/stream1"))

Experience the Full Ecosystem (v1 Launch Candidate)

Run the integrated engine and tactical dashboard in a single command:

python run_sentinel_v1.py

Access the dashboard at http://localhost:8000

Mission-Critical Performance

Sentinel Core is engineered for high-frequency processing. Below are typical throughput metrics (FPS) achieved on standard institutional hardware:

Layer	Hardware Substrate	Resolution	Throughput
Detection (YOLO)	Apple M2 Max	1080p	~85 FPS
Detection (YOLO)	NVIDIA RTX 4090	1080p	~140 FPS
Semantic Search	Apple M2 Max	1080p	~12 FPS
Semantic Search	NVIDIA RTX 4090	1080p	~25 FPS

Note: Benchmarks are internal and based on v0.6.0 optimizations.

Verify Performance Locally

Run the institutional benchmarking utility to measure performance on your specific silicon:

python benchmark.py

Progressive Evolution

Sentinel Core is a living framework. We push technical updates and architectural refinements as they are validated through our ongoing strategic deployments. Our goal is to maintain a continuous stream of progress:

• v0.1.0 (Feb 2026): Core Schema & AI Abstraction Interfaces.
• v0.2.0 (Feb 2026): Functional Stream Processing & Native Intelligence wrappers (YOLO).
• v0.2.1 (Feb 2026): Sovereign Semantic Search via CLIP embeddings.
• v0.4.0 (Feb 2026): Strategic Response (Alerting) & Evidence Persistence (Recording).
• v0.5.0 (Feb 2026): Silicon-Aware Hardware Acceleration (CUDA/MPS).
• v0.6.0 (Feb 2026): Sovereign Memory (Persistent SQLite Archive).
• v0.8.0 (Feb 2026): Sovereign Intelligence Dashboard & Semantic Search Bridge.
• v1.0.0 (Target): Official Launch — Zero-Latency Surveillance Orchestration.

---

Philosophy

Surveillance is not just about recording video; it's about autonomous situational awareness. Sentinel Core provides the schemas that transform raw pixels into strategic data.

© 2026 AB Labs • Strategic Systems Engineering.