NetPulse AI

About

NetPulse is an open-source control layer for infrastructure automation.

It turns a user request into a plan, checks risk, applies policy, asks for approval when needed, executes through adapters, verifies the result, and saves an audit record.

The first real working domain is Cisco network operations. The architecture is designed to extend to servers, firewalls, storage systems, monitoring tools, lab systems, and other infrastructure platforms through adapters.

Workflow

intent -> plan -> risk -> policy -> approval -> execute -> verify -> audit

The point is not to let AI freely run commands. The point is controlled execution with proof.

Features

• Structured execution plans before actions run.
• Risk classification for each request.
• Policy checks before execution.
• Approval gate for write or risky actions.
• No arbitrary CLI execution.
• Adapter-based execution model.
• Post-action verification.
• JSON audit artifacts for every request.

Current use cases

• Check VLANs, trunks, interfaces, and device status on Cisco network devices.
• Run approved Cisco network operations.
• Use chat interfaces such as Telegram to ask questions about the network.
• Generate a plan and audit trail for infrastructure requests.
• Run simulated demo workflows for compute, storage, and lab-system readiness.

Chatbot interface

NetPulse can be used from chat tools such as Telegram. The chatbot is not a general CLI shell. It routes requests through fixed intents, policy checks, approval gates, verification, and audit logging.

What you can do from chat

• Ask what VLANs exist on a switch.
• Check trunk status between switches.
• Find where a MAC address is connected.
• Check interface status and descriptions.
• Run approved read-only troubleshooting commands.
• Request a safe change, such as adding a VLAN, with approval required before execution.
• Receive a short answer in chat with the result and audit reference.

Example chat flow

User: Check trunk status on SW-CORE-01
NetPulse: Plan generated. Risk: READ_ONLY. Running approved check.
NetPulse: Trunk status found. Result saved to audit log.

User: Add VLAN 120 to SW-ACC-01
NetPulse: Plan generated. Risk: LOW_CHANGE. Approval required before execution.

Genesis-style demo

This demo is simulated. It does not require real servers, routers, firewalls, storage, or lab devices.

It shows how NetPulse would prepare a lab or research environment before a job runs by checking mock network, compute, storage, and instrument systems. The goal is to demonstrate the control flow: plan, risk check, execution, verification, and audit.

In production, the same adapter model could connect to real infrastructure such as routers, switches, firewalls, Linux servers, storage systems, monitoring tools, APIs, or lab instruments.

How real checks would work

User request
  -> NetPulse planner
  -> risk and policy check
  -> adapter selection
  -> SSH/API/device check
  -> parsed result
  -> verification
  -> audit record

Examples of future real adapters

• cisco_ios.py for switches and routers.
• paloalto.py or fortigate.py for firewalls.
• linux_ssh.py for Linux servers.
• windows_winrm.py for Windows servers.
• prometheus.py or zabbix.py for monitoring systems.
• storage_api.py for storage platforms.
• kubernetes.py for clusters.

Run the demo

Normal readiness check:

python3 demos/genesis_style/run_demo.py

Dry run:

python3 demos/genesis_style/run_demo.py --dry-run

Simulated write blocked by approval:

python3 demos/genesis_style/run_demo.py --simulate-write

Simulated write with approval:

python3 demos/genesis_style/run_demo.py --simulate-write --approve

Why it exists

AI can generate commands. That is not enough.

Infrastructure automation needs planning, safety rules, approvals, verification, and audit records. NetPulse focuses on that control layer.