text stringlengths 0 92 |
|---|
# ============================================================================== |
# RCS-Nexus Core Dependencies |
# ============================================================================== |
# Asynchronous & Event-Driven Engine |
asyncio>=3.4.3; python_version < "3.8" # Core async routing for multi-agent loops |
# Data Serialization & Validation |
pydantic>=2.5.0 # High-performance data validation for signal bridges |
# Knowledge Graph & Memory Management |
networkx>=3.1 # Graph structures for the semantic memory manager |
# LLM & Context Integration Utilities |
openai>=1.3.0 # Standard interface for context evaluation (optional endpoint) |
tiktoken>=0.5.1 # Token counting for precise context window control during sleep |
# Testing & Quality Assurance |
pytest>=7.4.0 # Framework for local unit testing |
pytest-asyncio>=0.21.1 # Support for testing asynchronous agent loops |
# Development & Environment Management |
python-dotenv>=1.0.0 # Secure management of project environment variables |
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
RCS-Nexus: Reactive Control System & Context Optimization Framework
RCS-Nexus is an open-source, production-ready system architecture and modular framework designed for orchestrating multi-agent AI ecosystems and optimizing large language model (LLM) context windows.
By utilizing a decentralized 7-layer Reactive Control System (RCS) topology alongside automated memory defragmentation routines, RCS-Nexus prevents state explosion and significantly lowers token costs in long-running autonomous workflows.
🔑 Core Features
- Decentralized Multi-Agent Orchestration: Asynchronous pipeline dividing tasks into specialized operational nodes (Temporal, Spatial, Sensory, and Processing streams).
- Hierarchical Memory Management: Segregates incoming data into short-term execution buffers and a persistent semantic graph structure.
- Signal Translation Bridge: High-throughput translation layer that normalizes raw telemetry, positional variables, and unstructured real-time inputs into formal vector tokens.
- Context Defragmentation (Sleep Routine): Automated background cycles that halt external state consumption to analyze memory graphs, consolidate historical records into abstract schemas, and flush execution buffers.
📐 System Architecture
The ecosystem relies on an asynchronous event-driven layout structured as follows:
RCS-Nexus/
├── rcs_core/
│ ├── agents/
│ │ ├── temporal.py # Timeline management & epoch validation
│ │ ├── spatial.py # Relational coordinate calculations
│ │ ├── sensory.py # Raw ingest & telemetric capture
│ │ └── processing.py # Local decision engines & state evaluations
│ ├── memory/
│ │ ├── graph_manager.py # Semantic knowledge graph graph construction
│ │ └── buffer_cleaner.py # Context defragmentation & sleep executor
│ └── translators/
│ └── signal_bridge.py # Structural JSON serialization
The Data Flow Topology
[Raw External Data] ──> (Sensory Node) ──> [Signal Bridge] │ ▼ [Temporal Log] <─── [Orchestrator Matrix] ───> [Spatial Coordinate Node] │ ▼ [Processing / Decision] │ ┌───────────────────────┴───────────────────────┐ ▼ ▼ [Short-Term Buffer] [Memory Graph Manager] │ ▲ └─────────────> (Sleep Routine Cycle) ──────────┘ [Defragment & Compress]
--- ## 🛠️ Deep Dive: The Defragmentation (Sleep) Cycle In long-lived autonomous applications, direct state accumulation results in contextual drift and linear token cost escalation. RCS-Nexus solves this via an isolated system routing called the Sleep Cycle: 1. Isolation: The main orchestrator temporarily buffers external incoming telemetry. 2. Analysis: The buffer_cleaner parses the active Short-Term Buffer to generate entity-relation clusters. 3. Consolidation: Intersecting data nodes are pushed to the underlying Knowledge Graph. Redundant structural noise is pruned. 4. Abstract Compression: Historical operational logs are condensed into static "state definitions", significantly lessening context weight for subsequent evaluations. 5. Flush: The ephemeral processing workspace is wiped, returning the active execution frame to base usage levels. --- ## 🚀 Quick Start ### Prerequisites * Python 3.10 or higher * Virtual environment manager (venv or conda) ### Installation Clone the repository and install the framework dependencies: bash git clone https://github.com cd RCS-Nexus python -m venv venv source venv/bin/activate # On Windows use: venv\Scripts\activate pip install -r requirements.txt ### Basic Usage Example python from rcs_core.orchestrator import NexusOrchestrator # Initialize the central matrix orchestrator = NexusOrchestrator(config_path="config.json") # Ingest an external raw signal stream raw_signal = {"timestamp": 1791641340, "vector_type": "spatial_node", "payload": {"x": 40.0, "y": 1.85}} orchestrator.process_signal(raw_signal) # Explicitly trigger the defragmentation (sleep) routine to optimize context orchestrator.trigger_sleep_cycle() --- ## 📄 License This project is licensed under the MIT License - see the LICENSE file for details. Any modifications, commercial applications, and closed-source private deployments are permitted under the condition that copyright notices are preserved.
- Downloads last month
- 76