Gain insight into your users’ Gemini for Google Cloud usage with GCP Cloud Logging and Monitoring feature.
Gemini for Google Cloud is your own AI-powered helper, from Code Assist, which helps your developers write better code more quickly, to Cloud Assist, which helps cloud operators manage consumption more effectively.
But until today, it has been difficult to pinpoint exactly how your internal users are use Gemini.
Google announced that Gemini for Google Cloud now supports GCP cloud logging and monitoring. While Cloud Monitoring provides aggregated data on 1-day, 7-day, and 28-day Gemini for Google Cloud active users and response counts, GCP Cloud Logging, which is currently in public preview, logs requests and answers between individual users and Gemini for Google Cloud.
GCP Cloud Logging
Logs are helpful in a few situations in addition to providing clients with broad insight about Gemini’s impact:
- To monitor where your AI-generated content comes from
- To document and examine Gemini for Google Cloud user usage
When activated, this feature, which is optional, logs your users’ Gemini activity for Google Cloud to Cloud Logging (charges apply).
Log entries are created for every request to and response from Gemini for Google Cloud once it is activated. An entry like the one shown below would be provided by Logs Explorer in a normal request entry:
Gemini for Google Cloud
This entry should be noted for a number of reasons:
- Details about the request are contained in the text of jsonPayload. In this instance, the request was to finish the Python code using the input def fibonacci.
- The labels include information about the product (code_assist), the method (CompleteCode), and the user who made the request (cal@google.com).
- The instance, location, and resource container (usually project) where the request took place are all indicated by the resource labels.
A normal response entry will look like this:
It is possible to identify request and response pairs since the request_ids inside the label are the same for this pair of requests and responses.
In addition to the Log Explorer, Log Analytics lets you evaluate your log data with questions like “How many requests did User XYZ make to Code Assist?”
Cloud Monitoring
Gemini for Google Cloud monitoring metrics assist you in addressing enquiries such as:
- In the last day or seven days, how many distinct active customers have utilised Gemini for Google Cloud services?
- In the last six hours, how many responses did my users receive overall from Gemini for Google Cloud services?
Anyone using a Gemini for Google Cloud product can leverage Cloud Monitoring support, which records answers and active users as Cloud Monitoring metrics that can be used to construct dashboards and alerts.
You can utilise these metrics as part of Cloud Monitoring dashboards since they are available with Cloud Monitoring. When Gemini for Google Cloud usage is identified, a “Gemini for Google Cloud” dashboard is automatically installed under “GCP Dashboards”:
Another option for examining metrics and applying filters to obtain a more comprehensive picture of your usage is Metrics Explorer. To accomplish this, choose the Metrics Explorer’s “Cloud AI Companion Instance” active resource:
Response_count, which can be filtered for Gemini Code Assist or the Gemini for Google Cloud method (code completion/generation), is the number of responses sent by Gemini for Google Cloud in the example above.
What comes next?
Google Cloud is always adding new features to these existing ones, and it is especially working on improving Code Assist logging and metrics to provide even more visibility and understanding of how you utilize Gemini Code Assist and its effects.
