Trisul AI 2025: Conversational AI for Next-Gen Network Monitoring

Lets be real, network monitoring isn’t getting easier. Logs are scattered across multiple systems, dashboards are buried under layers of menus, and reports are scattered in different formats. Somewhere in that mess the actually important stuff is hiding and urgent issues aren’t going to wait for you to learn the tools. To get to the stuff that you really want in the format you desire is a matter of knowing all the menus, forms, options, and settings.  In the age of LLMs, these seem outdated.

We are introducing an alpha version of our Conversational AI feature. Think of it like ChatGPT, but for your network. Instead of digging through menus, logs and reports hoping to stumble on answers you just ask. The system gets what you mean, pulls together the right info, and serves up actionable insights faster. It uses our knowledge base and entire documentation via RAG to help get you the data.

Behind the scenes, Trisul AI runs on Retrieval-Augmented Generation (RAG) and a MPC server and tools.

This is a conversational AI system that lets users query network analytics data using natural language. It combines:

• A Model, currently we support Gemini AI (Google’s LLM) for understanding queries. You just need a key.
• MCP (Model Context Protocol) for tool execution – we have created a MCP Servers with a number of tools which use the low level Trisul TRP API.
• Trisul Network Analytics tools (via ZMQ) for data retrieval. The TRP API allows structured queries
• RAG (Retrieval-Augmented Generation) for knowledge lookups. All of Trisul Network Analytics User Guides, API docs, Wiki and Knowledge Base are chunked and stored as vectors in Chroma DB, a database designed for fast similarity searches.

“Our immediate vision for Trisul Network Analytics is to bring the increasingly familiar conversational AI interface to our users.”

Vivek R, CTO, Unleash Networks

The best part? You get to talk with your network, not at it.

Forget CLI commands, dropdowns, and dashboard rabbit holes. You just ask.

“What’s the traffic trend on this interface in the last 24 hours?”
• “Which ASNs pushed the most traffic this week?”
• “Any weird spike in inbound flows lately?”

That’s it!

Trisul AI handles the rest. Including data retrieval, real time lookups, and response generation, all behind the scenes. You get clear answers in plain language, or if you prefer visuals, it can throw out tables, graphs, evena PDF report for sharing.

Speaking of tables and charts, a quick shoutout to Partha from our Dev Team. Terminal screens weren’t exactly designed to handle fancy tables and charts especially across different shells and OS quirks. He wrestled with width limits, scrollable datasets, and all the weird edge cases that make developers swear. Thanks to his developer stubbornness quality, you now get clean readable insights right inside the terminal. No extra clicks, no detours.

Here is an example.  Notice how it uses the knowledge queries from RAG to answer informal and sometimes imprecise queries.  It remembers context from previous interactions to know that the user typically wants last 5 minutes as the timeframe.

> what are top ips on airtel wan on ho fortigate

🤖 (Bot): Here are the top 5 hosts by traffic on the Airtel WAN at the head office firewall in last 5 mins
+---------------------+----------------+ 
| Host IP             | Traffic (MB)   | 
+---------------------+----------------+ 
| 10.25.30.151        | 242.29 MB      | 
| 10.26.12.104        | 227.34 MB      | 
+---------------------+----------------+

The more you use it, the sharper it gets. Trisul AI adapts to your query style and focus areas, so the insights feel tuned to how you work. And if youre the tinkering type you can switch to your own choice of model to shape the behavior of AI.

Setting up Trisul AI is simpler than you might expect. Think of it as preparing a dedicated workspace for your AI assistant. Clean, organized, and ready to help you explore your network analytics effortlessly.

First, make sure your system is ready. On Debian/Ubuntu, a quick update and installation of Python’s package manager and virtual environment tools is all you need:

These commands ensure your system has the foundation for a smooth, isolated setup.

Next, create your dedicated workspace by a virtual environment and step inside it:

By activating this environment, you’re giving Trisul AI its own space to work, free from conflicts with other projects.

Now, bring your AI assistant to life by installing the Trisul AI CLI:

And finally, launch it:

No menus to hunt through, no commands to memorize. Just a smooth, guided way to access traffic trends, peering insights, DDoS alerts, and more.

This is merely a sneak peek into our initial experimental work bringing conversational AI into Trisul Network Analytics.

Stay tuned for more about this as our developers improve this flow.