Thanks for sharing this! I’ve been exploring AI in networking and see a lot of potential. Many network engineers still think it’s just "hype" with no real applications. Mention AI or ChatGPT in Reddit’s networking threads, and it often gets downvoted. I’m working on a Tier 1 NOC agent, and it’s been very effective. It handles tickets, runs commands, and updates tickets on its own. You can even create specific instructions, and it will follow them consistently. With an API integration to an NMS, it can analyze historical data, generate summaries, and even reply to customer emails in the future with more polish. It removes the need for manual triage! Even without RAG, it works well. Adding a simple RAG setup, like a Cisco command guide as a vector store, boosted its performance significantly. Huge potential here!
Hi John, Its really nice one. I have a doubt for Ex. if command outputs that we are getting as a response, if its larger, it would easily max out the tokens, Instead of directly sending the entire command outputs to LLM, Can we store those outputs in ChromaDb & retrieve it based on our User Input that we have given?
@@johncapobianco2527 Quite common to have to RDP onto a customer's Windoze jump server (with restricted permissions) and then a Windoze management server desktop to use PuTTy on a customer network in UK. Hopefully that will change : )
Hey Bro , have done several iterations over my take on ai networking, your use case is very nice, would love to chat and collaborate if possible. Would love to learn from a wise one.
Happy to see someone showing more interesting use cases
Thanks for sharing this! I’ve been exploring AI in networking and see a lot of potential. Many network engineers still think it’s just "hype" with no real applications. Mention AI or ChatGPT in Reddit’s networking threads, and it often gets downvoted.
I’m working on a Tier 1 NOC agent, and it’s been very effective. It handles tickets, runs commands, and updates tickets on its own. You can even create specific instructions, and it will follow them consistently.
With an API integration to an NMS, it can analyze historical data, generate summaries, and even reply to customer emails in the future with more polish. It removes the need for manual triage!
Even without RAG, it works well. Adding a simple RAG setup, like a Cisco command guide as a vector store, boosted its performance significantly. Huge potential here!
amazing what the agent can do , can you at some point create a tutorial how we can do this . very fascinating !
Hi John, Its really nice one. I have a doubt for Ex. if command outputs that we are getting as a response, if its larger, it would easily max out the tokens, Instead of directly sending the entire command outputs to LLM, Can we store those outputs in ChromaDb & retrieve it based on our User Input that we have given?
Very inspiring. Thanks for sharing! I love these walkthroughs videos! Keep it up!
My favourite part of this video was commiserating with someone who also has a needy dog 😂
super cool. thanks for sharing.
Working on something similar. Thank you for sharing🎉
How beefy do the servers need to be hosting ollama and the AI bits?
@@wiz0rdsworld733 this is a very low end home PC with an old RTX 2060
This is great, well done and thanks for sharing. Will you be present in the Cisco Live this year ?
@@michaelmichel1871 non presenting but hopefully attending
This is great - one observation, a managed services provider(which accounts for a lot) might struggle to do this on a secure customer network?
I'm not sure why? The LLM is local and private and confidential - there are no cloud components here
@@johncapobianco2527 Quite common to have to RDP onto a customer's Windoze jump server (with restricted permissions) and then a Windoze management server desktop to use PuTTy on a customer network in UK. Hopefully that will change : )
Possibily dumb question, but nonetheless, who will plug Ethernet cables?
@@korvinkori_ drones .. robots … outsourced humans
what about any agent that can help me in game development? like in unity for example? agent that can see my screen, voice and has memory
It's 2025 and I have to ask somebody else to reset the clock on my car😂
Twice a year😢
Hey Bro , have done several iterations over my take on ai networking, your use case is very nice, would love to chat and collaborate if possible. Would love to learn from a wise one.
Hi