AI Semantic Monitor

From detection goals to anomaly events

Describe your concern.
Let AI check your observability data.

Which errors keep recurring? Which endpoints are slower than usual? Include the objects, time window and clues in your detection prompt. When an anomaly is found, review the analysis in the event and continue investigating.

01 Define goals in natural language Describe what to check and what you want to understand.
02 Connect the data context Add metric sets and log sources to ground the analysis.
03 Cover different monitoring needs Start with examples for stability, performance, security and cost.
04 Analyze and respond continuously Set the detection frequency and alert policy, then follow up on findings.

Turn your investigation approach
into an AI detection goal

Describe the data to examine, the time window and signs of anomalies, without first splitting the question into multiple queries and conditions. Specify the impact, evidence or troubleshooting advice you need, then set how often AI revisits the question.

View documentation

Turn your investigation approachinto an AI detection goal

Choose the data objects
that guide the analysis

Alongside the detection prompt, select metric sets or log sources by log index. These objects become the model’s context for understanding your services, resources and business concerns.

View documentation

Start with a good question
for your routine checks

Built-in prompts cover stability, security, performance and cost. Recurring errors, unusual logins, changes in endpoint latency and rising resource costs can all be detection goals. Adapt the objects and criteria to your environment.

View documentation

Start with a good questionfor your routine checks

Connect ongoing analysis
to your incident response

Configure alert policies and permissions so anomaly events enter your existing workflow. Review AI findings in the event details and investigate with observability data. Set a credit budget per check, and test saved prompts to review execution progress and results.

View documentation

Frequently asked questions

Where do I create an AI monitor?

In the Guance console, open Monitoring > AI Monitor and create a monitor. Enter its name, detection frequency, problem category and prompt. Add data objects as needed, then configure alert policies and permissions.

How do I make the detection goal specific?

State the objects, time window, clues and expected output. For example, examine order-service error logs from the last hour, identify recurring anomalies, and explain affected endpoints, supporting evidence and suggested next steps. Adapt built-in examples to your data.

How does this differ from using AI to generate detection rules?

AI Semantic Monitor repeatedly analyzes data according to your goal and schedule. Natural-language rule generation helps configure detection rules. Choose the approach that fits how you express the problem and what you need to monitor.

Can testing a prompt create real events?

Yes. Tests run against saved configurations and may generate events and send alert notifications. Check the data scope and alert policy before testing, then refine the prompt using execution progress and results.

How are analysis costs and results managed?

Advanced settings provide a credit budget per check, with credits deducted for actual usage. At the budget limit, further queries stop and results use the data already collected; actual usage may exceed the configured budget. Review analysis and alert notifications in the event details.

Describe your concern.Let AI check your observability data.

Where do I create an AI monitor?

How do I make the detection goal specific?

How does this differ from using AI to generate detection rules?

Can testing a prompt create real events?

How are analysis costs and results managed?

Related capabilities and documentation

Describe your concern.
Let AI check your observability data.