What’s New: OPAL Co-Pilot! (Powered by GPT)

By Jeremy Burton,April 27, 2023

 No longer are users divided between those who are “point-and-click” and those who can code. With OPAL Co-Pilot, everyone can code.

OPAL, Before AI

Back in the early days of Observe, we had big debates internally about whether we wanted to build a purely visual product that users could point and click their way to a root cause, or whether we would offer a console window and a scripting language. We couldn’t agree, so we figured the right thing to do was implement both.

To make it easier for a user to learn our query language — OPAL (Observe Processing & Analytics Language) — we decided each UI click should generate OPAL, which we would keep track of in a hidden console window. Then, if a user wanted to learn OPAL, they could pull up the console window and see the code their clicks had generated.

OPAL, Powered by GPT

With the advent of Large Language Models (LLMs) and GPT, we thought we could do much better. Rather than having new users click around in the UI and see what OPAL is generated, why not simply allow them to pull up the OPAL console and say what they want to do in a natural language? For example

// Extract the IP address from the log column

And for more advanced users, some queries can’t be expressed through the UI.

// Calculate the average response time for each unique IP address, then filter requests with a 200 status code, and then sort the results in descending order

To do all this, we use our LLM — trained on OPAL — to generate statements and execute them. The user experience is quite seamless, simply add instructions to a commented line and then hit CMD-K. Observe will then return a maximum of three OPAL statements. The user clicks on the desired statement (or select it using <TAB>) and it’s then posted into the console and executed. The following video shows a simple example of extracting, counting, sorting, and charting IP addresses.

OPAL, Easier For Everyone

We believe this is a first in the world of Observability — a co-pilot which can help engineers explore logs and metrics using only natural language as an input. New or novice users can now learn how to advance their skills in a very familiar and natural way, and experienced users who are grappling with the exact syntax can quickly re-jog their memory to get back on track. No longer are users divided between those who are “point-and-click” and those who can code. With OPAL Co-Pilot, everyone can code.

And, like O11y GPT, the best bit? It’s available in production today!

Ready to revolutionize your observability practice? Click here to unlock the benefits of our latest GPT-related integrations like OPAL CO-Pilot (and more) for your team!