Cloudera Docs

Tool execution

The Agent Studio sandbox environment manages file access through two distinct directories, Artifact File Directory and Workflow Data Directory, that are designed to separate writable outputs from read-only inputs.

Artifact File Directory

Artifact File Directory is the designated and sole writable location within the sandboxed environment for tool execution.
Artifact File Directory (/workspace) has the following key functions:
  1. Current Working Directory (CWD) - Is the default CWD for all tool executions.
  2. Output Destination – Specifies the location where tool outputs are stored.
  3. UI Visibility – This directory is visible within the Agent Studio UI, allowing you to view and download artifacts.
Practical Implications:
  • The sandbox automatically configures /workspace as the working directory on tool execution.
  • Tools writing files using a relative path, such as output.csv places them in /workspace.
  • Files generated here are immediately accessible in the Agent Studio UI.
  • The path to this directory is also accessible to the tools using the SESSION_DIRECTORY environment variable.

Workflow Data Directory

The Workflow Data Directory is a critical subdirectory within the Workflow directory, designed to centralize input datasets, runtime artifacts required by tools, and other project files. Users with Contributor access or higher in the Agent Studio project can access this directory.

The Workflow Data Directory is mounted inside the sandbox environment in the /workflow_data directory. The WORKFLOW_DATA_DIRECTORY environment variable also points to the /workflow_datadirectory.

Workflow Data Directory has the following functionality and key features:
  • Standardized Access - The directory provides a consistent, path-independent method for accessing project files.
  • Read-Only - The directory is mounted as read-only, meaning you can read files from it, for example, input files, configuration files, or data files, but you cannot write to it.
  • Content - The directory contains all uploaded and created files from your project that are available to the workflow.