Start Here
Welcome to Sedai! We are excited to help your team start exploring opportunities with your cloud resources. Use this guide to get your team up and running quickly.
Last updated
Was this helpful?
Welcome to Sedai! We are excited to help your team start exploring opportunities with your cloud resources. Use this guide to get your team up and running quickly.
Last updated
Was this helpful?
Onboarding requires permission to access your as well as a to inform Sedai's analysis and build models. Sedai uses Identity and Access Management (IAM) to securely connect to cloud accounts without an agent, however for Kubernetes clusters we recommend integrating via Sedai's .
Heads up: You likely will need help from your security team to give Sedai sufficient access — we recommend raising a ticket as soon as possible. Learn more about Sedai's , , and when you integrate your data.
Sedai is thoughtfully engineered to help your team save time by safely acting on your behalf. The system will only act when given permission, which is controlled via three different modes: Datapilot, Copilot, and Autopilot ().
While you control permissions within Sedai, these permissions are ultimately defined during the integration process. You can choose to provide either read-only or read-write access:
Read-only: Allows Sedai to analyze monitoring data, however the system can only predict low-confidence opportunities for cost/performance savings and flag potential availability issues. With read-only access, Sedai can only operate in Datapilot mode. It's important to note that read-only access significantly limits Sedai's features and capabilities. We recommend a Crawl-Walk-Run approach () to build confidence with the system and gradually allow Sedai to make changes to your resources.
Read-write: Allows Sedai to build advanced models based on reinforcement learning for maximum savings; the system will ask for your permission to make changes on your behalf. With read-write access, Sedai can operate in any mode. It's important to note that by default all integrations run in Datapilot mode, which simulates read-only access. Sedai will ask for permission to act in Copilot mode, or act autonomously in Autopilot mode. You can update modes from the Settings page ().
We recommend connecting dev resources first to explore Sedai and its different modes. However, keep in mind that dev resources typically have low usage and traffic, both of which inform Sedai's machine learning (ML) models. As a result, Sedai's opportunities, recommendations and savings for these types of resources will not be significant. To experience Sedai's full capabilities and see realistic savings, we recommend integrating a production account or cluster once you're ready.
Each cloud account or Kubernetes cluster integration requires read-only access to at least one monitoring data source in order to learn and analyze resource behavior. Sedai automatically imports the Golden Signals: latency, traffic, errors, and saturation.
Sedai builds monthly, weekly, and daily seasonality models for each of your resources. When you integrate monitoring data, the system will enhance its predictions over the course of 2 to 3 weeks. Alternatively, you can backfill metric data to expedite the learning process.
For Sedai to be most effective, it needs sufficient traffic and high quality metric data.
Select from the list below to view connection requirements and access details:
By default, Sedai leverages tags to accurately infer metrics and map them to your infrastructure. The system pre-populates standard tag values based on industry best practices for naming conventions. However, if your organization configured custom tags you will need to provide the exact tag value used — .
