Software observability platform Honeycomb today announced the launch of Query Assistant, a new capability leveraging generative AI to enable natural language querying for observability. The company claims it is the first observability platform to provide this feature, which scales its query power and makes observability more usable for all engineering levels.

The Query Assistant capability has been developed using OpenAI models and empowers engineers to ask questions in plain English instead of a query language, allowing them to get fast feedback on their code performance and behavior without having advanced knowledge of query-based programming languages like SQL. 

“We believe the best developer tools are the ones that get out of people’s way so that all of their energy can be focused on the problem they are trying to solve,” Charity Majors, cofounder and CTO at Honeycomb, told VentureBeat. “Our Query Assistant lets you leapfrog past all that time and effort spent figuring out what to query, structuring your query understanding your schema, and simply tell us in plain English what you’re looking for.”

Honeycomb said that the feature is a distinctly different approach to AI compared to traditional application performance monitoring (APM) tools that apply AI to data analytics for features like automated alerting. Through this generative AI capability, the platform allows users to build queries that can be further modified and used to obtain relevant data for further analysis. 

This announcement follows closely on the heels of the company’s successful $50 million series D funding round. The Query Assistant will be available to all Honeycomb users as an experimental feature that teams can turn off. In addition, the company clarified that the platform does not share user data with OpenAI and will neither retain any user data for training models.

Making observability more accessible

Honeycomb’s flagship product revolves around its robust query builder, which serves as the bedrock of its platform. The latest addition, Query Assistant, aims to further enhance query builders’ access to diverse data types and analysis graphs for all users. The company believes that observability in the hands of everyone on a team will allow every engineer to understand the effects of their code, which will also allow them to write better code in the first place. 

“Query Assistant makes us the first observability platform with a fully-executing natural language querying experience, where you can just ask for what you want in plain English, and the tool will run the query and return the right results,” Majors told VentureBeat. “After that, you can use the normal query builder to iterate on the generated results and tweak the query until you find what you’re looking for.” 

Majors said the company has always been committed to a friendly, intuitive user interface and set out to build a tool that is usable by anyone, from experts to new hires. She said the new feature would democratize observability and empower developers of all backgrounds by eliminating the need for highly technical language.

“Users shouldn’t have to burn mental cycles on wrestling with complex development tools,” she said. “Even with our UI, the challenge has often been orientation: Enabling a user to translate the thought in their brain into the fields available for them to query upon. With Query Assistant, even that step is smoothed, as we can map the English language a user uses, to the wide array of custom fields captured in their data — and translate it directly to a query run in our UI.”

The company has historically been cautious about using AI in alerting and automation workflows due to the potential for increased risk and variance, said Majors.

“The data in this world isn’t regular or predictable enough to rely on AI to find trends or signals, either — and finally, both false positives and false negatives are painful and expensive for your team,” she said. “But we’ve always been curious about how machine analysis can pair with human intuition to speed it up rather than try to replace it.”

The need for natural language querying

The company recognized the need for an intuitive system that enables non-technical team members to participate in the development process after observing new engineers describe the debugging process of experienced engineers as “magical leaps of intuition.”

“Access to powerful insights about your software has historically been limited by the engineer’s ability to memorize custom query languages, interact with complex UIs or magically know all the field names to interact with,” said Majors. “Not every engineer is a specialist in these areas — nor should they have to be. But every engineer should understand the code they just created and be able to ask questions about it.”

According to Majors, while composing a query from scratch can be challenging and time-consuming, tweaking a nearly-correct query is something that even non-technical team members can easily do.

“With Query Assistant, regardless if you are on day two of a new job using an unfamiliar system or you’re trying Honeycomb for the first time, users of all levels can describe or ask things in natural language, and the tool will generate a relevant Honeycomb query for you to tweak, iterate on or share with your team,” she explained. 

Unique pricing model

Honeycomb’s pricing model sets it apart from competitors in the observability market, according to the company. Unlike other providers, Honeycomb does not charge additional fees per service, host, memory, custom field or seat. This unique approach allows teams to scale their observability efforts without incurring hidden or unexpected costs.

“It’s hard to believe that you could be charged by CPU or the number of instances in 2023,” said Majors. “When you force your users to delete their most valuable bits of telemetry for cost reasons, you are not only exposing the fact that your storage system is archaic, you are making it your users’ problem.”

She continued: “Our pricing encourages you to query as much as you want because we want you to understand your systems. Users can pack as many key/value pairs (“custom metrics”) as they want into wide events, as richer data makes telemetry more powerful.”

According to the company, Query Assistant currently only supports query generation. However, they plan to introduce features that will enable the modification of existing queries within the system in the near future.

“Query Assistant might give less accurate results for large schemas,” said Majors. “For now, we want to focus on creating new queries accurately. This won’t impact most users because the majority of teams lack datasets that are large enough to make a noticeable difference in accuracy.”

A future of opportunities with AI 

According to Majors, the company’s philosophy centers around leveraging the strengths of machines and humans alike. Machines are exceptional at processing data and performing calculations, while humans excel at assigning meaning to information. By combining the unique skills of both machines and humans, the company aims to offer insights and drive innovation in the tech industry.

“Any machine can process data and tell you whether or not a spike occurred; only a person can tell you whether that spike was good or bad, expected or not,” she said. “So we are excited about the possibility of AI to help surface those insights — helping you observe how a power user interacts with the part of the system they know most intimately and use that information to be a power user too.” 

The company encourages users to try the new feature and help improve the system through feedback. 

“The most important information about the sociotechnical systems we work on lives in people’s heads,” Majors added. “We want people to try it and give us lots of feedback. Since it’s complementary to Query Builder, if you’re finding that it doesn’t do what you need, you can still use Query Builder.”