Diagram: System Overview
Relevant source files
The following files were used as context for generating this wiki page:
The AI Network Operations Platform is a self-hosted assistant designed to function as an AI Network Engineer. It enables network operations teams to manage multi-vendor environments through natural language, automating tasks such as infrastructure discovery, troubleshooting, and configuration management while ensuring every change is human-approved.
Sources: README.md:3-8, CLAUDE.md:3-12
Overview
The platform solves common operational challenges by providing a secure, local-first environment where AI handles complex analysis and data correlation. It supports a wide range of vendors and protocols to ensure compatibility across the entire enterprise stack.
Key Philosophy
- Local-First & Self-Hosted: The platform runs within your own data center or cloud environment, ensuring data privacy.
- Human-in-the-Loop: No changes are applied to the network without explicit approval from a human engineer.
- Auditability: Every decision made by the AI is explained and logged for full transparency.
Sources: README.md:3-8, CLAUDE.md:36-43
Specialized Team Members (Agents)
Instead of thinking of the system as complex software, think of it as a team of specialized engineers (agents). When you ask a question, the Master Architect determines which specialist is best suited to help:
| Agent | Responsibility |
|---|---|
| Discovery Agent | Performs inventory enumeration (scanning and listing what exists). |
| Troubleshooting Agent | Diagnoses live faults, missing routes, or traffic drops. |
| DDI Agent | Manages DNS, DHCP, and IP addresses. |
| Security Agent | Audits firewall policies and identifies security posture risks. |
| Configuration Agent | Analyzes configuration drift and compliance against baselines. |
| Documentation Agent | Generates diagrams, runbooks, and inventory reports. |
| Packet Analysis Agent | Queries and summarizes finished packet captures (pcaps). |
| Automation Agent | The sole executor that carries out approved change requests. |
| Consultant Agent | Asks clarifying questions if your request is too vague. |
Sources: backend/app/llm/prompts/init.py:186-278, CLAUDE.md:27-34
Supported Vendors and Protocols
The platform uses a standardized connector system for each vendor. It normalizes data from various sources (SSH, SNMP, APIs) into a consistent format so that the AI can reason about a Cisco router and a Juniper switch using the same logic.
- Networking: Cisco (IOS, IOS-XE, NX-OS), Juniper (JunOS), Arista (EOS), Palo Alto (PAN-OS), Fortinet (FortiOS), F5 (BIG-IP).
- DDI: BlueCat, Infoblox, Route53.
- Cloud/Virt: AWS, Azure, VMware.
Sources: CLAUDE.md:14-25, README.md:23-28
Architecture
The platform consists of seven core components working together to provide the AI engineering experience.
- Frontend: The operations console where engineers chat with the AI, view the topology map, and approve changes.
- API: The central brain that handles requests, manages security, and coordinates agent sessions.
- Worker: The background engine that runs long-running discovery jobs, takes config backups, and handles packet analysis without blocking your session.
- Postgres: The system of record that stores long-term data like device inventory, user accounts, and change logs.
- Neo4j: A living topology map that tracks how every device, interface, subnet, and DNS record connects to everything else.
- Redis: A high-speed messenger used to manage the background job queue and temporary data.
- Ollama: The local inference engine that runs the AI models themselves, keeping all intelligence local to your network.
Sources: README.md:12-21, deploy/docker/docker-compose.yml:21-168
Deployment Options
- Docker Compose: Ideal for lab environments or small teams. It runs the entire stack on a single server.
- Kubernetes (Helm): Designed for production. It provides higher availability and better resource management across multiple nodes.
Sources: deploy/docker/README.md:3-11, deploy/kubernetes/netops/README.md:3-14
Diagrams
System Overview
This diagram shows how users interact with the platform and how the platform connects to your infrastructure and external services.
The diagram illustrates the flow from the user interface through the core platform components to the various managed network families.
Sources: CLAUDE.md:14-25, README.md:12-21
Agent Routing Flow
When you send a question, it follows this path to ensure the right specialist handles it.
This flow shows the Master Architect's role in dispatching tasks and the safety mechanism that diverts changes to an approval gate.
Sources: backend/app/llm/prompts/init.py:186-350, backend/app/agents/ddi/agent.py:11-23
Change Approval Lifecycle
Every modification to the network follows a strict "Human-in-the-Loop" lifecycle.
The lifecycle ensures that no AI-generated configuration is ever deployed without manual verification and a safety backup.
Sources: backend/app/agents/ddi/agent.py:11-23, backend/app/agents/security/agent.py:12-26
Topology Data Model
This shows "what the platform knows" about your network, represented as a graph of relationships.
The topology model tracks logical and physical connections, allowing the AI to understand dependencies across the infrastructure.
Sources: CLAUDE.md:52-57, backend/tests/agents/documentation/test_documentation_agent.py:108-132
Conclusion
The AI Network Operations Platform provides a structured, safe, and multi-vendor approach to network automation. By combining specialized AI agents with a strict approval workflow and a deep understanding of network topology, it enables operations teams to move faster while maintaining high standards of security and reliability.
Sources: README.md:3-8, CLAUDE.md:36-43