AI-Inferred Service Descriptions

A key component of your service catalog are the descriptions that provide an easy to consume snapshot of what a service is for and its architecture, however these descriptions are often overlooked while catalogs are populated. OpsLevel uses the power of large language models to review documentation and structure within your code to generate useful descriptions.


Trust at OpsLevel

While AI and LLMs are exciting technologies, we understand how important it is to understand where your data is being used and stored. You can learn more about the specifics of this feature in this guide, but we also encourage you to review our security page to learn more about what we do to keep your information safe and secure.

How it works

When enabled on an account, this feature uses an intelligent agent to analyze the contents of your services’ code repositories. The agent will read any repository content; it prefers documentation, but it will also read code.

None of your code is stored permanently. The agent transiently scans your documentation and code and uses it to make an inference about what a service’s description should be. Similar to our repo checks, after computing a result, the original input code and documentation is eventually discarded.

Recognizing an inferred description

To help you distinguish AI-generated content from human-generated, we've added a robot-face emoji prefix to your description. If a user modifies the description, even by saving the inferred description, the identifier will be removed.

What other third-parties are used to analyze/process my data?

In addition to OpsLevel’s typical code-scanning functionality, we’re using a large-language model hosted by OpenAI.

If this feature is enabled, your repository contents could be submitted to the OpenAI API for analysis. (So by opting in to using this feature, OpenAI will be added as a data subprocessor of OpsLevel.)

We care deeply about who has access to your code and where it is stored. Per OpenAI API Data Usage Policies, OpenAI will not permanently store data submitted via their API or use it to train their models.

How do I get access?

The feature is currently disabled by default on accounts that were created before October 12th, 2023. Early Access to the feature can be turned on (or off) at any time through a request to your OpsLevel customer success team.

What's next?

This is OpsLevel's first foray into using large language models to help you build your catalog, but we're keen to leverage this more to meet our goals of making it easier for you to keep your catalog complete, accurate and rich without all the hassle. If you have ideas for what we should do next or questions about how we do things today, please reach out to us at [email protected].

What’s Next

Want to learn more about other ways we help you keep your catalog up to date without all the fuss?