<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: khurram javed</title>
    <description>The latest articles on DEV Community by khurram javed (@kj121).</description>
    <link>https://dev.to/kj121</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4251678%2Fd6f58da9-33f5-424f-a70b-c1ab3ac3b17f.png</url>
      <title>DEV Community: khurram javed</title>
      <link>https://dev.to/kj121</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/kj121"/>
    <language>en</language>
    <item>
      <title>Beyond Chatbots: Building a Zero-Trust AIOps Change Auditor for Cisco Networks with Local LLMs, NAPALM, and GNS3</title>
      <dc:creator>khurram javed</dc:creator>
      <pubDate>Thu, 19 Nov 2026 06:51:28 +0000</pubDate>
      <link>https://dev.to/kj121/beyond-chatbots-building-a-zero-trust-aiops-change-auditor-for-cisco-networks-with-local-llms-1ffb</link>
      <guid>https://dev.to/kj121/beyond-chatbots-building-a-zero-trust-aiops-change-auditor-for-cisco-networks-with-local-llms-1ffb</guid>
      <description>&lt;p&gt;Every network engineer knows the quiet tension of a 2:00 AM maintenance window.&lt;/p&gt;

&lt;p&gt;You have reviewed your Method of Procedure (MOP), secured change-advisory board (CAB) approval, and executed configuration updates across core routers. To verify, you run &lt;code&gt;show ip interface brief&lt;/code&gt;, confirm all interfaces are &lt;code&gt;up/up&lt;/code&gt;, and send a standard 5-packet ICMP ping across the transit link:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Sending 5, 100-byte ICMP Echos to 172.16.30.1, timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5)
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;With five clean exclamation marks, everything appears healthy. You close your terminal, update the change ticket to "Completed," and log off.&lt;/p&gt;

&lt;p&gt;Three hours later, morning business traffic ramps up, customer service lines light up, and the critical incident bridge convenes: &lt;strong&gt;enterprise web dashboards and database sessions between HQ and the Data Center are hanging and timing out.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;What went wrong?&lt;/p&gt;

&lt;p&gt;The control plane never dropped. OSPF remained in &lt;code&gt;FULL&lt;/code&gt; state. BGP remained &lt;code&gt;Established&lt;/code&gt;. Small 100-byte monitoring pings passed cleanly. However, an engineer had inadvertently committed &lt;code&gt;ip mtu 1400&lt;/code&gt; along with &lt;code&gt;no ip unreachables&lt;/code&gt; on a core transit link while masking the mismatch using &lt;code&gt;ip ospf mtu-ignore&lt;/code&gt;. &lt;/p&gt;

&lt;p&gt;As soon as real application traffic arrived with 1500-byte frames and the Don't Fragment (DF) bit set, the packets were silently dropped. Because ICMP unreachables were disabled, Path MTU Discovery (PMTUD) failed silently.&lt;/p&gt;

&lt;p&gt;Industry studies consistently reveal that &lt;strong&gt;over 70% of network outages are self-inflicted during scheduled maintenance windows&lt;/strong&gt;. The root cause is rarely a complete physical link collapse; rather, it is subtle &lt;strong&gt;"gray failures"&lt;/strong&gt; that easily slip past manual check sheets.&lt;/p&gt;

&lt;p&gt;In this article, I walk through how I designed and built &lt;strong&gt;&lt;code&gt;netops-ai-mop-auditor&lt;/code&gt;&lt;/strong&gt;, an open-source, zero-trust &lt;strong&gt;Pre/Post Maintenance Verification &amp;amp; AI Change Auditor&lt;/strong&gt; using &lt;strong&gt;Python, NAPALM, Netmiko, Flask, and local Llama 3.2 (via Ollama)&lt;/strong&gt;, validated end-to-end inside a &lt;strong&gt;GNS3&lt;/strong&gt; multi-AS Cisco IOS lab.&lt;/p&gt;




&lt;h2&gt;
  
  
  For Beginners: The Learning Path from Python Scripts to AIOps
&lt;/h2&gt;

&lt;p&gt;If you are new to network automation—setting up virtual Cisco labs or writing your first Python connection scripts—AIOps becomes intuitive once the underlying mechanics are clear.&lt;/p&gt;

&lt;p&gt;This project represents the practical evolution of a foundational multi-part automation series I previously published on LinkedIn. If you are starting from scratch or want to follow the progression step by step, start with these guides:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Step&lt;/th&gt;
&lt;th&gt;Topic / Focus&lt;/th&gt;
&lt;th&gt;LinkedIn Walkthrough&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;1. Lab Virtualization&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;GNS3 + VM Setup &amp;amp; Bridging&lt;/td&gt;
&lt;td&gt;
&lt;a href="https://www.linkedin.com/pulse/gns3-21-gui-vm-installation-integration-steps-khurram-javed" rel="noopener noreferrer"&gt;GNS3 2.1 GUI &amp;amp; VM Installation Steps&lt;/a&gt; — Setting up GNS3 and bridging virtual routers to your host machine via &lt;code&gt;Cloud1&lt;/code&gt; and &lt;code&gt;Hub1&lt;/code&gt;.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;2. Python CLI Basics&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Raw Socket Connections&lt;/td&gt;
&lt;td&gt;
&lt;a href="https://www.linkedin.com/pulse/network-automation-gns3-lab-telnet-cisco-router-using-khurram-javed" rel="noopener noreferrer"&gt;Network Automation: Telnet to Cisco Router&lt;/a&gt; — Understanding programmatic command execution and terminal handling.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;3. Secure Automation&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Enterprise SSH via Netmiko&lt;/td&gt;
&lt;td&gt;
&lt;a href="https://www.linkedin.com/pulse/network-automation-gns3-lab-ssh-cisco-router-using-netmiko-javed" rel="noopener noreferrer"&gt;Network Automation: SSH via Netmiko&lt;/a&gt; — Managing SSH sessions, handling command prompts, and navigating privilege levels.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;4. Multi-Vendor State&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Structured Config with NAPALM&lt;/td&gt;
&lt;td&gt;
&lt;a href="https://www.linkedin.com/pulse/network-automation-gns3-lab-configure-cisco-router-using-javed" rel="noopener noreferrer"&gt;Network Automation: Configure Router with NAPALM&lt;/a&gt; — Moving beyond screen-scraping toward structured multi-vendor getters and config management.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;5. Web Dashboards&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Telemetry via Nornir + Flask&lt;/td&gt;
&lt;td&gt;
&lt;a href="https://www.linkedin.com/pulse/network-automation-retrieve-devices-facts-interface-details-javed" rel="noopener noreferrer"&gt;Network Automation: Telemetry via Nornir &amp;amp; Flask&lt;/a&gt; — Surfacing live router metrics inside an interactive browser UI.&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Once you understand how to interact programmatically with network devices and render operational state on the web, the next engineering challenge is: &lt;strong&gt;How do we introduce Large Language Models (LLMs) into NetOps workflows safely without risking network downtime?&lt;/strong&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  Why Naive "LLM Chatbots" Fail in Production NetOps
&lt;/h2&gt;

&lt;p&gt;When engineers first experiment with AI, the common approach is to copy and paste hundreds of lines of &lt;code&gt;show running-config&lt;/code&gt; before and after a change into an LLM with a prompt like:&lt;/p&gt;

&lt;p&gt;&lt;em&gt;"Here is the configuration before and after my change. Did anything break?"&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;In real-world network operations, this naive approach fails due to two distinct engineering hazards:&lt;/p&gt;

&lt;h3&gt;
  
  
  1. The Hallucination &amp;amp; Arithmetic Hazard
&lt;/h3&gt;

&lt;p&gt;Small local models (such as 3B to 8B parameter models) are probabilistic language models, not deterministic calculation engines. When presented with 600 lines of unstructured Cisco IOS configuration text, they regularly miss subtle wildcard mask typos, miscalculate BGP prefix counts, or hallucinate non-existent interface names. Relying on an LLM to accurately diff routing tables in context memory is an architectural anti-pattern.&lt;/p&gt;

&lt;h3&gt;
  
  
  2. The Unacceptable Blast Radius of Autonomous Write Access
&lt;/h3&gt;

&lt;p&gt;Granting an AI model autonomous write access (&lt;code&gt;configure terminal&lt;/code&gt;) to enterprise core infrastructure is a risk no prudent engineering team will permit. A single prompt injection or hallucinated command syntax could trigger an unrecoverable routing loop or widespread blackout.&lt;/p&gt;

&lt;h3&gt;
  
  
  The Solution: The Anti-Hallucination Split Architecture
&lt;/h3&gt;

&lt;p&gt;To make AI production-grade in networking, we enforce a strict separation of concerns:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Deterministic Python Engine (&lt;code&gt;NAPALM&lt;/code&gt; + &lt;code&gt;Netmiko&lt;/code&gt; + &lt;code&gt;difflib&lt;/code&gt;):&lt;/strong&gt; Machines excel at structured data collection, cryptographic snapshot hashing, mathematical line-by-line diffing, and synthetic data-plane probes.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Local Semantic Auditor (&lt;code&gt;Llama 3.2&lt;/code&gt; via &lt;code&gt;Ollama&lt;/code&gt;):&lt;/strong&gt; LLMs excel at reasoning over structured context, auditing semantic scope compliance against authorized human change requests (MOPs), and writing human-readable Post-Implementation Review (PIR) reports.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8k1c0ycctviaiye3k85n.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8k1c0ycctviaiye3k85n.png" alt="System Architecture" width="800" height="1708"&gt;&lt;/a&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  The GNS3 Lab Topology: Multi-AS Enterprise Core
&lt;/h2&gt;

&lt;p&gt;The validation environment is hosted in &lt;strong&gt;GNS3&lt;/strong&gt; and simulates an enterprise-to-data-center transit architecture:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F0q85en3dsoqx0l8kxnqx.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F0q85en3dsoqx0l8kxnqx.png" alt="GNS3 Lab Topology" width="800" height="505"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;R1 (&lt;code&gt;AS 65001&lt;/code&gt; – HQ Edge):&lt;/strong&gt; Advertises HQ LAN &lt;code&gt;192.168.10.0/24&lt;/code&gt; into eBGP across &lt;code&gt;Fa0/0 (10.12.0.1/30)&lt;/code&gt; to R2.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;R2 (&lt;code&gt;AS 65002&lt;/code&gt; – Core ASBR):&lt;/strong&gt; Bridges eBGP (AS 65001) to the internal core. Runs OSPF Area 0 on &lt;code&gt;Fa2/0 (10.23.0.1/30)&lt;/code&gt; to R3 and peers via iBGP (&lt;code&gt;Loopback0: 2.2.2.2&lt;/code&gt; to &lt;code&gt;3.3.3.3&lt;/code&gt;) applying &lt;code&gt;next-hop-self&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;R3 (&lt;code&gt;AS 65002&lt;/code&gt; – DC Edge):&lt;/strong&gt; Advertises DC Server subnet &lt;code&gt;172.16.30.0/24&lt;/code&gt; into BGP.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Out-of-Band (OOB) Management Plane:&lt;/strong&gt; Interfaces &lt;code&gt;R1 e2/0 (172.20.40.211)&lt;/code&gt;, &lt;code&gt;R2 e1/0 (172.20.40.212)&lt;/code&gt;, and &lt;code&gt;R3 e1/0 (172.20.40.213)&lt;/code&gt; connect through &lt;code&gt;Hub1&lt;/code&gt; to host loopback adapter &lt;code&gt;Cloud1 (172.20.40.0/24)&lt;/code&gt;, completely isolated from data-plane traffic.&lt;/li&gt;
&lt;/ul&gt;




&lt;h2&gt;
  
  
  Zero-Trust Engineering Controls
&lt;/h2&gt;

&lt;p&gt;To guarantee operational safety, the platform implements three strict design controls:&lt;/p&gt;

&lt;h3&gt;
  
  
  1. The 4-Layer Snapshot Engine (&lt;code&gt;tools.py&lt;/code&gt;)
&lt;/h3&gt;

&lt;p&gt;Every Pre-Check and Post-Check captures four distinct operational layers across &lt;code&gt;R1&lt;/code&gt;, &lt;code&gt;R2&lt;/code&gt;, and &lt;code&gt;R3&lt;/code&gt;:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Configuration Layer:&lt;/strong&gt; Cleaned &lt;code&gt;running-config&lt;/code&gt; (volatile Cisco IOS byte-count headers and timestamps are stripped so clean diffs produce zero noise).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;IGP Layer:&lt;/strong&gt; OSPF neighbor Router-IDs and &lt;code&gt;FULL&lt;/code&gt; adjacency states.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;BGP Overlay Layer:&lt;/strong&gt; Peer session states, received prefix counts, and best-path (&lt;code&gt;*&amp;gt;&lt;/code&gt;) validity for critical HQ and DC prefixes.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Data Plane Layer:&lt;/strong&gt; Synthetic &lt;code&gt;100-byte&lt;/code&gt; reachability pings and progressive &lt;code&gt;1500-byte DF-bit&lt;/code&gt; Path MTU probes in both directions.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  2. Regex-Enforced Read-Only Guardrails (&lt;code&gt;guardrails.py&lt;/code&gt;)
&lt;/h3&gt;

&lt;p&gt;The AI agent has &lt;strong&gt;zero write tools&lt;/strong&gt;. Every CLI query requested by Llama 3.2 passes through strict Python validation that permits only &lt;code&gt;show&lt;/code&gt;, &lt;code&gt;ping&lt;/code&gt;, and &lt;code&gt;traceroute&lt;/code&gt; commands while blocking shell escapes (&lt;code&gt;;&lt;/code&gt;, &lt;code&gt;| redirect&lt;/code&gt;) and write verbs (&lt;code&gt;configure&lt;/code&gt;, &lt;code&gt;reload&lt;/code&gt;, &lt;code&gt;write&lt;/code&gt;, &lt;code&gt;erase&lt;/code&gt;) using word-boundary regex patterns.&lt;/p&gt;

&lt;h3&gt;
  
  
  3. 100% Air-Gapped Local Inference
&lt;/h3&gt;

&lt;p&gt;Running &lt;strong&gt;Llama 3.2&lt;/strong&gt; locally via &lt;strong&gt;Ollama&lt;/strong&gt; ensures that router configurations, internal IP schemes, and operational telemetry never leave your local management perimeter.&lt;/p&gt;




&lt;h2&gt;
  
  
  The 4-Stage Maintenance Lifecycle in Action
&lt;/h2&gt;

&lt;p&gt;The Flask web dashboard (featuring &lt;strong&gt;Dark/Light mode&lt;/strong&gt; and real-time Server-Sent Events streaming) guides the engineer through a strict 4-step workflow:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Select Maintenance Scenario (MOP) &amp;amp; Scope:&lt;/strong&gt; Pick a planned activity (&lt;code&gt;MOP-101&lt;/code&gt; link hygiene, &lt;code&gt;MOP-102&lt;/code&gt; new DC subnet onboarding, &lt;code&gt;MOP-103&lt;/code&gt; iBGP tuning, or &lt;code&gt;MOP-104&lt;/code&gt; custom manual GNS3 work) and define the Authorized Change Scope.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Run Pre-Checks (&lt;code&gt;PRE_ACTIVITY&lt;/code&gt;):&lt;/strong&gt; Lock the 4-layer baseline. Once verified, the UI displays a green &lt;strong&gt;CLEARED&lt;/strong&gt; banner authorizing the engineer to begin work.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Execute Activity (Manual GNS3 CLI or Simulation):&lt;/strong&gt; Make real changes directly inside GNS3 consoles—or test with the built-in &lt;strong&gt;Clean&lt;/strong&gt; vs. &lt;strong&gt;Misconfiguration&lt;/strong&gt; simulators.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Manual Engineer Confirmation → Post-Check &amp;amp; AI Verdict:&lt;/strong&gt; When finished, the engineer clicks &lt;strong&gt;Confirm Changes Done&lt;/strong&gt;. Python captures &lt;code&gt;POST_ACTIVITY&lt;/code&gt;, builds the &lt;strong&gt;Pre vs. Post Comparison Matrix&lt;/strong&gt; and &lt;strong&gt;Unified Config Diff&lt;/strong&gt;, and streams Llama 3.2's audit verdict.&lt;/li&gt;
&lt;/ol&gt;

&lt;h3&gt;
  
  
  Case A: Clean Maintenance Approved (&lt;code&gt;MOP-101&lt;/code&gt;)
&lt;/h3&gt;

&lt;p&gt;When only the authorized interface description on &lt;code&gt;R2 FastEthernet2/0&lt;/code&gt; is updated, the Comparison Matrix confirms zero regressions, the Config Diff highlights the exact 1-line change, and Llama 3.2 approves sign-off:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ffrbj3qd0dd5wjortn0km.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ffrbj3qd0dd5wjortn0km.png" alt="Approved MOP Run" width="800" height="530"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  Case B: Silent Misconfiguration Caught &amp;amp; Blocked
&lt;/h3&gt;

&lt;p&gt;When a simulated mistake introduces &lt;code&gt;ip mtu 1400&lt;/code&gt; + &lt;code&gt;no ip unreachables&lt;/code&gt; (or drops a BGP prefix / &lt;code&gt;next-hop-self&lt;/code&gt;), the &lt;code&gt;1500B DF-bit&lt;/code&gt; probe immediately flags &lt;code&gt;REGRESSION&lt;/code&gt;, the Unified Config Diff pinpoints the unauthorized lines, and Llama 3.2 rejects the change—activating a one-click rollback button:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fsl1i91f165fzsvddrhta.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fsl1i91f165fzsvddrhta.png" alt="Regression Caught" width="799" height="532"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  Audit-Ready Deliverables: 4-Sheet Excel &amp;amp; Markdown PIR
&lt;/h3&gt;

&lt;p&gt;For CAB reviews and compliance audits, clicking &lt;strong&gt;Export Excel (&lt;code&gt;.xlsx&lt;/code&gt;)&lt;/strong&gt; in the top-right header generates a styled 4-tab workbook (&lt;code&gt;Executive Summary&lt;/code&gt;, &lt;code&gt;Pre-Post Matrix&lt;/code&gt;, &lt;code&gt;Config Diffs&lt;/code&gt;, &lt;code&gt;AI Audit Trace&lt;/code&gt;), alongside a timestamped Markdown Post-Implementation Review (&lt;code&gt;reports/pir_*.md&lt;/code&gt;):&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fsnrie5cj4wen9j8b3v3b.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fsnrie5cj4wen9j8b3v3b.png" alt="Excel Audit Report" width="800" height="420"&gt;&lt;/a&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  Key Takeaways for SysOps &amp;amp; NetOps Engineers
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;1. Let Python do the math; let AI do the audit:&lt;/strong&gt; Never rely on an LLM to diff raw configs when deterministic libraries (&lt;code&gt;NAPALM&lt;/code&gt;, &lt;code&gt;difflib&lt;/code&gt;) do it with 100% mathematical accuracy.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;2. Enforce guardrails in code, not prompts:&lt;/strong&gt; Prompt instructions can be bypassed or misread; Python allowlists cannot.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;3. Audit trails drive adoption:&lt;/strong&gt; Producing a clean &lt;code&gt;.xlsx&lt;/code&gt; workbook and PIR Markdown file turns an experimental AI script into an operational tool CAB teams and managers can trust.&lt;/p&gt;




&lt;h2&gt;
  
  
  Try It Yourself (Open Source)
&lt;/h2&gt;

&lt;p&gt;The complete source code, GNS3 baseline configs (&lt;code&gt;R1.cfg&lt;/code&gt;–&lt;code&gt;R3.cfg&lt;/code&gt;), and setup instructions are available on GitHub under the &lt;strong&gt;GNU GPLv3&lt;/strong&gt; license:&lt;/p&gt;

&lt;p&gt;👉 &lt;strong&gt;&lt;a href="https://github.com/shahkh-eng/netops-ai-mop-auditor" rel="noopener noreferrer"&gt;github.com/shahkh-eng/netops-ai-mop-auditor&lt;/a&gt;&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Clone the repository, boot the 3-router topology in GNS3, pull &lt;code&gt;llama3.2&lt;/code&gt; via Ollama, and test your own manual CLI changes in GNS3 to watch the auditor evaluate your work in real time. Contributions, new MOP scenarios, and feedback are always welcome!&lt;/p&gt;




&lt;h2&gt;
  
  
  References &amp;amp; Further Reading
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Network Programmability and Automation (2nd Edition)&lt;/strong&gt; — Matt Oswalt, Christian Adell, Scott S. Lowe, and Jason Edelman (&lt;em&gt;O'Reilly Media&lt;/em&gt;, 2023). An essential reference for Python, NAPALM, Netmiko, and structured network operations.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Site Reliability Engineering: How Google Runs Production Systems&lt;/strong&gt; — Betsy Beyer, Chris Jones, Jennifer Petoff, and Niall Richard Murphy (&lt;em&gt;O'Reilly Media&lt;/em&gt; / &lt;a href="https://sre.google/books/" rel="noopener noreferrer"&gt;sre.google&lt;/a&gt;). Invaluable guidance on change management, post-incident reviews, and mitigating automation risk.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;ReAct: Synergizing Reasoning and Acting in Language Models&lt;/strong&gt; — Shunyu Yao et al. (&lt;a href="https://arxiv.org/abs/2210.03629" rel="noopener noreferrer"&gt;arXiv:2210.03629&lt;/a&gt;). The foundational academic research behind tool-calling reasoning loops in LLM agents.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;RFC 1191 — Path MTU Discovery&lt;/strong&gt; — J. Mogul and S. Deering (&lt;em&gt;IETF&lt;/em&gt;). Detailed breakdown of the ICMP Type 3 Code 4 fragmentation signaling mechanism.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;RFC 4271 — A Border Gateway Protocol 4 (BGP-4)&lt;/strong&gt; — Y. Rekhter, T. Li, and S. Hares (&lt;em&gt;IETF&lt;/em&gt;). The core standard defining iBGP next-hop resolution and route advertisement rules.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Cisco DevNet Code Exchange&lt;/strong&gt; — &lt;a href="https://developer.cisco.com/codeexchange/" rel="noopener noreferrer"&gt;developer.cisco.com/codeexchange&lt;/a&gt;. Community-driven repository for Cisco automation, APIs, and AI integrations.&lt;/li&gt;
&lt;/ol&gt;

</description>
      <category>networking</category>
      <category>aiops</category>
      <category>python</category>
      <category>devops</category>
    </item>
  </channel>
</rss>
