Environments & Operations#
Environments & Operations covers Model Training pages for shared runtimes, cloud accounts, power automation, and external integrations. Use it to give teams bookable environments, connect cloud spend sources, cut idle compute cost with schedules, and wire calendars or CI/CD into environment workflows.
Use this section when you need to:
- Share and book ML runtimes across teams without conflicts.
- Link cloud accounts used by environments and power schedules.
- Automatically start and stop cloud instances on a timetable.
- Connect Google Calendar or CI/CD to environment bookings.
From these pages you can see environment availability, configured cloud links, schedule coverage, and which external systems are connected.
Next: start with Shared Environments, add Cloud Connections for cloud-backed resources, then create Power Schedules and optional Integrations. For experiments and registries, see Experiment Tracking. Administrative actions require the Organization manager role.
Shared Environments#
Shared Environments is where administrators publish bookable runtimes and teams reserve them for training and testing. Use it to reduce contention and keep installed software and access consistent.
Use this page when you need to register an environment, check availability, book or release a slot, or temporarily deactivate bookings.
Next: create a shared environment, then book it from the table or calendar view.
Row actions
| Button | Action |
|---|---|
| Book the environment for a selected time range. | |
| Release a reservation and make the environment available again. | |
| Temporarily disable bookings for the environment. | |
| Remove the environment from the organization catalog. |
Create a shared environment#
To register a shared environment:
- Open Model training → Shared Environments.
- Click + ADD.
- Enter a Name and select a Resource type.
- Configure optional properties such as SSH access, Description, IP, and Software.
- Click CREATE.
New environments appear in the table together with status, booking, and software information.
Cloud Connections#
Cloud Connections links cloud provider accounts that Environments & Operations uses for resources and power automation. Use it when you need a new cloud account available for shared environments or Power Schedules.
Next: click + ADD to run the connection wizard for a supported provider, then assign resources to schedules or environments as needed.
Power Schedules#
Power Schedules automates cloud instance start and stop across AWS, Azure, Google Cloud, and Alibaba Cloud so idle resources are not left running. Each cloud resource can belong to only one schedule; a schedule may include multiple daily power-on and power-off triggers.
Use this page when you need to cut idle compute cost, enforce off-hours power policies, or review which instances are covered by automation.
Next: confirm cloud permissions, create a power schedule, then assign instances.
Use + ADD to create a schedule, and use the Actions column to activate, deactivate, or delete existing schedules.
Use
to activate a schedule.
Use
to deactivate a schedule.
Use
to delete a schedule.
Cloud permissions#
Cloud resources require start and stop permissions before schedules can operate.
| Provider | Permissions |
|---|---|
| AWS | ec2:StartInstances, ec2:StopInstances |
| Azure | Microsoft.Compute/virtualMachines/start/action, Microsoft.Compute/virtualMachines/deallocate/action |
| Alibaba Cloud | ecs:StartInstance, ecs:StopInstance |
| Google Cloud | compute.instances.start, compute.instances.stop |
Schedule detail page#
Select a schedule Name to open the detail view.
Summary cards display:
- last execution time
- assigned resource count
- configured time zone
- validity period
A visual 24-hour timeline displays powered-on and powered-off intervals throughout the day.
Use the Instances tab to manage assigned resources and the Triggers tab to review configured schedules.
Create a power schedule#
To create a schedule:
- Open Model training → Power Schedules.
- Click + ADD.
- Enter a schedule name and select a time zone.
- Configure validity dates and triggers.
- Click CREATE.
Manage schedule resources#
To assign resources to a schedule:
- Open the schedule detail page.
- Go to the Instances tab.
- Select ADD INSTANCES TO SCHEDULE.
- Choose instances and click ADD.
To remove resources:
- Select instances in the Instances tab.
- Click REMOVE INSTANCES FROM SCHEDULE.
Edit or delete schedules#
Use the schedule detail page to:
- update triggers
- modify validity dates
- activate or deactivate schedules
- delete schedules
Changes are applied after selecting EDIT or confirming the delete operation.
Integrations#
Integrations connects external scheduling and CI/CD systems to Shared Environment workflows so bookings and pipeline events stay in sync with OptScale AI.
Use this page when you need calendar-driven reservations or CI/CD hooks for environment lifecycle events.
Next: configure Google Calendar or a CI/CD integration after Shared Environments are in place.
Supported integrations may include:
- Google Calendar
- GitLab
- GitHub
- Jenkins
The page uses a card layout where each integration contains setup guidance and operational details.
Google Calendar#
The Google Calendar integration synchronizes Shared Environment reservations with a shared calendar view.
| Element | Purpose |
|---|---|
| Description | Explains how booking intervals appear in the calendar. |
| Status | Displays whether a calendar is connected. |
| CONNECT CALENDAR | Starts the calendar connection flow. |
CI/CD integrations#
GitLab, GitHub, and Jenkins integrations allow CI/CD pipelines to report environment status and deployment metadata to OptScale AI.
| Integration | Typical use |
|---|---|
| GitLab | Deployment pipeline stage. |
| GitHub | GitHub Actions workflow step. |
| Jenkins | Post-build or deployment stage. |
Pipeline examples use $ENV_COLLECTOR_URL to send deployment status and environment metadata updates.
See also#
- Model Training — Model Training menu overview.
- Core Services — AI Gateway, policies, and FinOps.
- Architecture Overview — Platform structure and request flow.