Workflows in Prospector Studio let you compose multi-step automated processes with conditional logic, approval gates, and artifact passing between steps.
Creating a Workflow
From the Studio UI, navigate to the Workflows section to create and manage workflows. Each workflow is defined as a directed graph of tasks.
Task Types
Workflows support several task types:
| Task Type | Description |
|---|---|
| Container | Run a Docker container with custom commands |
| Python | Execute Python scripts |
| SQL | Run database queries |
| LLM | Invoke an LLM with a prompt and context |
| Branch | Conditional routing based on output |
| Gate | Require human approval before proceeding |
| Array | Iterate over a collection |
Workflow Features
Conditional Branching
Route execution based on task outputs. Branch tasks evaluate conditions and direct the workflow down different paths.
Approval Gates
Insert human-in-the-loop checkpoints. Gate tasks pause execution until an authorized user approves or rejects the step.
Artifacts
Tasks can produce artifacts (files, data) that downstream tasks consume. Artifacts are managed automatically and available throughout the workflow execution.
Real-Time Monitoring
Monitor workflow executions in real-time through the Studio UI. Each task shows its status, output, and timing information as the workflow runs.
Execution
Workflows can be triggered:
- Manually from the Studio UI
- By an agent during a conversation
- Via API through the GraphQL endpoint