Environments
Environments are runtime targets where workflows actually execute.
The platform UI is the control plane. Environments are the execution plane.
This separation enables centralized design and governance while supporting distributed execution across cloud and edge devices.
Environment Types
Cloud Environment (Default)
Each project includes a default cloud environment.
Use it for:
- early development
- quick validation
- centrally managed execution
Custom Environments (Edge / On-Prem)
Custom environments are user-managed runtime targets.
Use them for:
- low-latency processing near data source
- restricted-network/on-prem requirements
- hardware-specific execution (GPU devices, gateways)
Environment Lifecycle
- Create environment record in project
- Install and run runtime agent with environment ID
- Validate status and heartbeat
- Deploy workflows to environment
- Monitor health and execution behavior
- Update, restart, or retire environment
Status Model
Common status states:
online: healthy and reporting heartbeatsoffline: heartbeat missing beyond threshold
Practical implication:
- Deployments to offline targets should be blocked or delayed
- Monitoring should alert on repeated online/offline transitions
What to Validate Before Deployment
- Environment status is online
- Required models are available and compatible
- Datasource connectivity is valid from target location
- Bucket paths and permissions are accessible
- Secrets and endpoint configuration are present
Do not deploy critical workflows to an environment that has not passed heartbeat and dependency validation checks.
Operational Best Practices
- Use explicit environment naming by location and purpose
- Separate dev, staging, and production targets
- Keep one agent identity per physical runtime target
- Track environment-level incidents in operations logs
- Apply controlled rollout when deploying major workflow updates
Security Practices
- Use scoped credentials per environment
- Rotate tokens and secrets regularly
- Restrict network access to required upstream/downstream endpoints
- Minimize exposed management surfaces on edge nodes
Troubleshooting Priorities
When an environment fails, check in this order:
- Agent/container process health
- Network path to platform APIs
- Credential/token validity
- Runtime resource pressure (CPU, RAM, storage)
- Workflow/model compatibility issues
Environment Reference Pages
- Environment Setup and Agent Install
- Heartbeat and Status Monitoring
- Workflow Deployment to Environments
- Environment Troubleshooting