AI-Assisted Network Automation

Course map

What the modules cover

Free foundation track

The free track has 10 lessons in 2 modules, free with an account.

1. 01From Your First Script to Network RecordsFree · 5 lessons

   Start with text and printed output, then learn numbers, lists, and functions before working with loops and network records. Browser checks use changed inputs to verify the skills you practice.

   • Your First Python Script: Text and Output
   • Numbers and Lists from Network Data
   • Functions: Name a Reusable Step
   • Conditionals and Loops
   • Dictionaries, Sets, Files, and Exceptions

   Open these lessons
2. 02Parsers, Objects, and a PackageFree · 5 lessons

   Extend your core Python with regular expressions, device objects, and libraries. Assemble parsers, audit checks, and a report into the netaudit package; local command-line practice is an extension.

   • Regular Expressions for Show Output
   • Classes and Objects, Part 1
   • Classes and Objects, Part 2
   • Libraries, pip, and Virtual Environments
   • Modules, Packages, and the Capstone

   Open these lessons

The paid course

The paid course has 35 lessons in 10 modules, and 32 of those lessons are graded. Its modules are numbered 0 through 9, so Start Here is module 0 of the paid course.

1. 00Start Here2 lessons

   What the course promises, and how the browser-first path works. Network FDE here means you own the whole change, not just the script. That runs from the first ticket through to proof and handoff. You see how the pages and labs work, when a local tool helps, and what this course cannot reproduce.

   • How This Course Works
   • The Network FDE Role
2. 01Just Enough Python5 lessons

   Learn the Python the rest of the course uses, on network records and saved output. A placement check lets an experienced Python reader skip ahead. The first graded record keeps a staff report separate from a measured fact. After that the module covers files, JSON, functions, and test output.

   • Can You Skip This Module?
   • Python: First Contact
   • Data Shapes: Lists, Dicts, and JSON
   • Loops, Decisions, and Functions
   • Your Working Environment
3. 02The Day-One Automation Loop4 lessons

   You run your first complete delivery loop on a made-up district office network. First you sort the reported complaints from the measured facts. Then you write the morning authentication problem down with a baseline and a target. You crawl LLDP, generate a topology you can review, and push a firewall policy change. You validate it deployed without breaking anything else. Then you hand the follow-up ticket to the help desk lead. Every required check runs in the browser.

   • Discover and Document a Network from One IP
   • Push Config from Data, and Prove It Deployed
   • Validate a Change Before and After
   • Side Quest: Mermaid Network Diagrams as CodeFree
4. 03Keep a Human in the Loop7 lessons

   Check every claim against what the devices report. Then route each failure to the role that can act on it. The module covers secret handling, reading a traceback, data-shape errors, logic bugs, diagnosing the environment, verification, and undo. You also sort the work into code, data, environment, process, training, staffing, and vendor, then record the next owner.

   • Secrets, Credentials, and Safe AI Usage
   • Read the Traceback, Find the Real Line
   • Judge the Shape: Dicts, Nested Dicts, and JSON
   • Spot the Bug You Didn't Write
   • Is It the Code or the Environment?
   • The Verification Habit (and Git as Undo)
   • Side Quest - Bug Swarm: The Human-in-the-Loop ArcadeFree
5. 04Build With AI3 lessons

   Write down what the tool must do before you ask a model for code. That means the problem it solves, its inputs and output, and what happens on error. It also means how you will accept it, and what stays out of version one. Then you check the code that comes back. You decide which actions can run on their own, which need approval, and which stay human work.

   • Working with an AI Assistant: the Loop
   • Vibe-Code a Tool with AI, and Verify It
   • Token Economics: Run Your Assistant Like a Network
6. 05Source of Truth and Git2 lessons

   Turn measured network state into a source of truth. It gets an owner, a seven-day freshness limit, and a line between what was observed and what was approved. Git keeps the history. You compare a later discovery with the record, trace the changes, restore an earlier standard, and hand freshness over to an owner.

   • Feed What You Discover into a Source of Truth
   • Version the Source of Truth in Git
7. 06Device APIs3 lessons

   Read state over RESTCONF, stream telemetry with gNMI, and make transactional changes over NETCONF. Then compare the three on access, data quality, licensing, maintenance window approval, rollback, and shared dependencies. You pick one and say why.

   • Talk to Devices over REST and RESTCONF
   • Stream Telemetry with gNMI
   • Change Safely with NETCONF and Confirmed Commit
8. 07From Scripts to Tools4 lessons

   Turn a working script into a tool another operator can run and support. Write down the operator, the maintainer, and the support boundary. Write down who has adopted it, the work trace, and what version one leaves out. Then structure the repository, write tests, make a repeat run safe, and return failure information somebody can use.

   • Project Structure for Automation Tools
   • Testing Beyond the Lab
   • Idempotency as a Discipline
   • Error Handling for Automation
9. 08Run It for Real3 lessons

   Plan and record an approved production change. You separate the acceptance checks from the operational outcomes. You settle approval and rollback before the window opens, and you test new requests against the scope. The rollout goes canary first, and you keep a redacted run record. The handoff includes the runbook, the support boundary, and when to measure again.

   • Change Workflow: Dry Run to Rollback
   • Logging, Run Records, and Audit Trails
   • Package It: Docker for Automation Tools
10. 09The Read-First Agent2 lessons

    You finish the course on a warehouse transfer task you run on your own, with a read-first agent. You decide what the problem is and what the tool may do. You decide how the change gets approved and rolled back, and who owns it after you. The labs grade those decisions.

    • The Read-First Agent
    • ChatOps: Human Approval for Agent Writes