Alert creation and debugging using AI

At Oodle AI, we have been tinkering with AI in the space of Observability. I would like to present our recent work of using LLMs for alert creation (translation from english prompts to PromQL queries) and in alert debugging where we combine statistics on timeseries data along with AI to provide blast radius and potential root cause to users.

In terms of rough outline of the demo, I would show:

1. AI-enabled alert creation experience.
2. Code walkthrough of how we are achieving this using Langchain and providing it relevant tools to deal with metric data.
3. Alert debugging experience where we apply statistics to find correlating metrics and use LLMs to summarize the findings.
4. Code walkthrough of LLM summarize for alerts.