Overview
TensorStax offers two execution modes to match your workflow preferences and project requirements. Choose between full autonomous execution or human-in-the-loop mode for greater control over the agent’s actions.Mode Options
Human-in-the-Loop Mode
Human-in-the-Loop Mode
Interactive Control: The agent asks for confirmation after every plan step and before the session begins.
- Review and approve each step before execution
- Request plan updates before the agent starts
- Redo any step as many times as needed
- Move to the next step only when you’re satisfied
Even in human-in-the-loop mode, TensorStax will automatically attempt to fix dbt model errors on its own before asking for your input.
Autonomous Mode
Autonomous Mode
End-to-End Execution: The agent executes the complete plan without interruptions.
- Runs the entire plan from start to finish
- Self-corrects when errors occur
- Automatically retries failed steps
- Continues until the expected result is achieved
Post-Execution Editing
Post-Execution Editing
Flexible Corrections: Regardless of execution mode, you can always make changes afterward.
- Use Quick Edit for simple modifications
- Access Open Editor View for comprehensive changes
- Manual file editing is always available
- Make adjustments even after autonomous completion
Choosing the Right Mode
Use Human-in-the-Loop When:
- Working on critical production models
- Learning how TensorStax approaches problems
- Need to review each step for compliance
- Want to customize the approach mid-execution
Use Autonomous When:
- Prototyping or exploring data
- Working on non-critical development tasks
- Trust the agent’s decision-making
- Need fast iteration and results
Best Practice: Start with human-in-the-loop mode for new projects to understand TensorStax’s approach, then switch to autonomous mode for similar tasks once you’re comfortable with the patterns.