Search Hook: CogniPath IoA Gateway integrates with Cisco Outshift’s AGNTCY Internet of Agents, enabling ACP-over-CogniPath routing, OASF directory registration, and LM-driven path selection for IoA-compatible multi-agent workflows.
Intent-Driven Networking Powered by Distributed Language Model Agents
CogniPath™ is a next-generation intent-driven packet networking system that leverages distributed language model (LM) agents to dynamically determine packet routing paths in real time. Instead of static configuration or legacy routing convergence delays, CogniPath™ embeds intent directly in packet metadata, enabling an intelligent, adaptive network that reacts instantly to topology changes, failures, and evolving policy requirements.
Unlike traditional routing protocols (BGP, OSPF, EIGRP), CogniPath™ dynamically adapts to the actual intent of the packet using on-demand inference, enabling:
- Adaptive routing based on live topology and telemetry.
- Real-time rerouting around failures or degraded links.
- Compatibility with legacy protocols via a protocol-agnostic interface.
Version: v0.1.0 Initial Public Pre-release (Public Architecture Preview)
See the Release Notes for details.
- Each packet carries embedded intent in its header/payload.
- Nodes act as autonomous LM agents interpreting this intent.
- Eliminates reliance on a single centralized controller.
- Distributed LMs act as decision agents to agree on optimal paths.
- Security hardened by token-hashing, UUID replay prevention, and consensus thresholds.
- Ensures deterministic decision outcomes in distributed settings.
- Moves decision-making to the edge control plane.
- Routing decisions are triggered dynamically by packets rather than static policies.
- Enables continuous optimization as topology evolves.
- Next‑generation replacement for spanning tree.
- Keeps all links active while preventing loops in real time.
- Dynamically adapts switching paths based on LM-driven network intent and live health telemetry.
- CogniPath’s architecture is designed for wireline and wireless environments, including LoRaMesh, Wi‑Fi 6, and 5G.
- Intent‑driven routing seamlessly adapts to variable wireless link quality and mobility conditions.
- Wireless capability is enabled through the same LM‑driven control plane, ensuring link resilience and real‑time rerouting without protocol‑specific dependencies.
- LM Prompt Adaptation: Mutates LM inputs dynamically (e.g., excludes failed nodes) based on telemetry.
- LM Response Auditability: Logs LM decisions in signed, traceable formats (
logs/lm_decisions.log). - Protocol-Agnostic Compatibility Layer: Translates LM-generated paths to legacy protocol equivalents (e.g., OSPF/BGP route injection).
- Adaptive Switching: Introduces LM-driven loop prevention and multi-path utilization for switching fabrics.
- Enterprise-Grade Switch Integration Demo
- Deployment of CogniPath-core in containerized environments (e.g., Arista EOS container-manager, Cisco IOx, Junos OS Evolved, etc.) on high-performance enterprise switches.
- Showcases LM-driven routing operating in parallel with traditional control planes.
- Illustrates intent-aware routing capabilities on enterprise hardware without exposing proprietary mechanisms.
CogniPath/
│
├── config/ # Configuration files
│ └── lm_roles.yaml # Defines LM roles: CogniCore™, CogniEdge™, CogniLite™
│
├── core/ # Public-safe core architecture
│ ├── README.md # Core overview (public-safe)
│ ├── __init__.py # Package initialization
│ ├── agent.py # LM orchestration layer (sanitized)
│ ├── node.py # Node logic abstraction (sanitized)
│ ├── packet.py # Packet structure & public-safe processing
│ ├── tokenlib.py # Public-safe token validation
│ └── utils.py # Utility functions
│
├── docs/ # Documentation
│ └── README.md # High-level technical overview
│
├── legacy_compat/ # Legacy routing protocol stubs (no LM logic)
│ ├── bgp/
│ │ └── parser.py # BGP compatibility stub
│ ├── eigrp/
│ │ └── parser.py # EIGRP compatibility stub
│ ├── ospf/
│ │ └── parser.py # OSPF compatibility stub
│ └── README.md # Legacy compatibility overview
│
├── logs/ # Logs & auditing (public-safe)
│ ├── README.md # Logging overview
│ ├── agent_audit.log # LM agent audit trail (public scrubbed)
│ └── lm_decisions.log # LM decision logs (public scrubbed)
│
├── tests/ # Public-safe test suite
│ ├── README.md # Overview of testing framework
│ ├── __init__.py # Package init for test discovery
│ ├── test_agent.py # Tests for agent abstraction layer
│ └── test_packet.py # Tests for packet structure
│
├── .gitattributes # Public repo IP boundaries & file handling
├── .gitignore # Ignored files for public repo safety
├── LICENSE.md # Proprietary license for CogniPath
├── NOTICE.md # IP protection notice & patent language
├── README.md # This file (overview & documentation)
├── main.py # Demo entry point for public build
└── requirements.txt # Public-safe dependencies
This public repository contains a limited, public-safe subset of CogniPath’s architecture.
Core intellectual property (LM internals, consensus algorithms, advanced prompt structures, and production routing logic) is maintained in the private CogniPath-core repository.
Patent protection is in process via USPTO Provisional Patent Application (PPA) covering:
- LM intent-embedded packet networking.
- Distributed LM consensus routing.
- Edge-triggered control plane inference.
- Adaptive Switching (loop-free multipath switching).
Clone the repository:
git clone https://github.com/<your-org>/CogniPath.git
cd CogniPathCreate a virtual environment & install dependencies:
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txtRun the demo:
python3 main.pyCogniPath™ supports interoperation with existing networks by translating LM-decided paths to formats understood by:
- OSPF
- BGP
- EIGRP
Note:
Legacy protocol translators in legacy_compat/ are public stubs. Full compatibility modules are maintained in private protected code.
This public repository is released under a restricted license.
Use is allowed for educational/research purposes.
Production/commercial usage without explicit permission is prohibited.
See NOTICE.md for detailed IP and licensing terms.
- Patent Status: PPA filed (USPTO)
- Private Repo: CogniPath-core (invite only)
- Public Roadmap:
- Phase 2 (Stable) – Core LM intent routing complete
- Phase 3 (In Development) – Expanded interoperability, observability, and edge readiness
- Phase 4 (Planned) – Advanced wireless adaptation, security, and enterprise integration
- Protocol Interoperability – Ensure compatibility with major routing protocols.
- Resilience Enhancements – Adaptive LM-driven routing that dynamically adjusts to real-time conditions.
- Performance Optimizations – Efficient packet management with enhanced telemetry and dynamic reroute.
- Scalable Management – Configurable policy modules and topology loaders for flexible deployments.
- Observability – Expanded logging, audit trails, and dashboard integration for operational insight.
- Wireless Network Adaptation – Mobility support across multiple wireless mesh technologies.
- Security & Trust – Strengthened replay protection, multi-agent consensus, and trust-weighted decision layers.
- LM Orchestration – Cooperative operation of different LM sizes for optimal performance.
- Enterprise Integration – Deployment in containerized environments on enterprise-class network hardware.
We welcome constructive collaboration, issue reports, and discussions.
Contact: Keenan Williams at [email protected]