Sand Agentic Workflows icon

Sand Agentic Workflows

Interact with Sand Agentic Workflows

Overview

This node, named "Sand Agentic Workflows," allows users to interact with Sand Agentic Workflows hosted on a Llama Deploy server. It is designed to send messages to specified workflows deployed on the server and optionally evaluate the results returned by those workflows. This node is useful in scenarios where automated or AI-driven workflows need to be triggered and their outputs processed within an n8n automation.

Practical examples include:

  • Sending user queries or commands to an AI-powered engineering assistant workflow and retrieving its response.
  • Triggering data processing or analysis workflows remotely and capturing their deliverables.
  • Integrating complex agentic workflows into larger automation pipelines for decision-making or content generation.

Properties

Name Meaning
Deployment Name The name of the deployment on the Llama Deploy server to target (default: "StateMachines").
Workflow Name The specific workflow on the deployment to invoke (default: "engineering_assistant").
Evaluate Result Boolean flag indicating whether to evaluate the result of the workflow ("true" or "false").
User Message The message string sent to the Llama Deploy server; can include expressions referencing input data.

Output

The node outputs an array of JSON objects, each corresponding to one input item processed. Each output JSON contains:

  • response: The extracted deliverable response from the workflow run result, or "No response available" if none found.
  • In case of failure, the output JSON includes:
    • success: false
    • error: A brief error message ("Failed to send message")
    • details: Detailed error information including the error message and the workflow name.

The node does not output binary data.

Dependencies

  • Requires connection credentials to the Llama Deploy server, including:
    • API server URL
    • Control plane URL
    • Optional MLflow tracking and AWS S3 credentials for experiment configuration
  • These credentials must be configured in n8n prior to using the node.
  • The node uses an internal client (LlamaDeployClient) to communicate with the Llama Deploy APIs.

Troubleshooting

  • Common issues:

    • Failure to connect to the Llama Deploy server due to incorrect URLs or network issues.
    • Authentication errors if API keys or credentials are missing or invalid.
    • Timeout errors if the workflow takes longer than the configured timeout (default 120 seconds).
    • Missing or malformed responses if the workflow does not return expected deliverables.

  • Error messages:

    • "Failed to send message": Indicates the node could not successfully invoke the workflow. Check credentials, network connectivity, and workflow availability.
    • "Failed to clean up session" (logged to console): Indicates an issue deleting the temporary session after execution; usually non-critical but should be investigated.

To resolve these errors:

  • Verify all credential fields and URLs.
  • Ensure the targeted workflow exists and is operational.
  • Increase timeout settings if workflows require more time.
  • Review server logs for detailed error diagnostics.

Links and References

Discussion