Tines 3B + some crazy ideas: Part 2
I’ve been continuing to see how far I can push something that started with just a handful of prompts.
In Part 2, the project starts getting a lot more interesting.
I added network troubleshooting agents that can select a site, discover the devices, run the commands they need, and turn the results into something an engineer can actually work with.
For example, I can select Boston, run an L1/L2 agent, and have it check the devices across that site. From there it can identify unreachable devices, inspect interfaces and SVIs, collect platform information, and summarize what it found.
I’ve also started building out:
- BGP troubleshooting agents
- Lab auditing
- Custom agent tasks
- Automated network-device upgrade workflows
- Activity and command tracking
- Token usage by agent run
- Full execution history, including who ran what and when
The upgrade piece actually came from a customer question. They have tens of thousands of devices and engineers still handling parts of the upgrade process manually. That immediately made me wonder: how much of that workflow could I build into this?
That’s really what this series is about.
Have an idea. Prompt it. Build it. Test it. Find the holes. Fix them. Then figure out what else it could do.
There are definitely things I already want to improve. For example, if authentication to a device fails, the agent shouldn’t continue trying every other command. Those are the kinds of guardrails that become obvious once you actually build and run this stuff.
Here’s Part 2:
Curious what the Tines community would add next. If you could hand a network or infrastructure problem to an agent, what would you want it to troubleshoot?