Resource Categories
1. Storage & Artifacts
Plugins can interact with the platform’s Liquid Data storage tiers.2. Model Providers
Access the unified AI model interface directly from your plugin nodes.3. Execution Context
Every plugin execution receives a richRemoteExecutionContext object. This object provides access to the current execution state, configuration, and credentials.
Context Schema
Thecontext object (of type RemoteExecutionContext) contains the following key attributes and methods:
| Attribute / Method | Type | Description |
|---|---|---|
plugin_config | dict | The effective configuration for the plugin, including admin and workspace overrides. |
integration_credentials | dict[str, dict] | A dictionary of credentials for the integrations required by the node. |
get_integration_credentials(name) | method | Helper method to safely retrieve credentials for a specific integration by its name. |
Best Practices
Storage usage
Storage usage
Treat storage as external I/O. Use deterministic paths and explicit retention behavior. Always use the platform’s file handling utilities to ensure compatibility across different storage backends.
Model usage
Model usage
Prefer provider-agnostic contracts. Keep provider-specific assumptions behind plugin adapters to allow users to switch models easily.
Context usage
Context usage
Pass only required context to reduce coupling and accidental data exposure. Use the context object to ensure your plugin respects workspace-level boundaries.
Failure handling
Failure handling
Use retries with backoff for transient external failures and return actionable errors. The platform’s execution engine will handle retries based on your node’s configuration.
Creating Integrations
Build robust connectors to external systems using platform resources.