Result validation is a built-in quality control feature in elvex flows that acts as a supervisor to ensure each step completes successfully before proceeding to the next step.
What is Result Validation?
Result validation is a background supervisor system that automatically checks whether each step in your flow is working correctly and producing valid outputs. The validator is enabled by default for every step you create in a flow.
Think of the validator as a quality control manager that reviews each step's output and determines whether it meets the expected criteria before allowing the flow to continue.
How Result Validation Works
When a step runs in your flow, the validation system:
Analyzes the step's output - Reviews what the step produced
Checks for completion - Ensures the step finished its intended task
Validates quality - Determines if the output meets expected standards
Makes a decision - Either approves the step to continue or flags it as failed
If validation passes, the flow moves on to the next step. If validation fails, the entire flow stops at that point, preventing potentially incorrect or incomplete data from propagating through subsequent steps.
When to Use Result Validation
Keep Validation Enabled When:
Accuracy is critical - When incorrect outputs could cause significant problems downstream
Complex processing - When steps involve complex analysis or data transformation
Multi-step dependencies - When later steps rely heavily on the accuracy of earlier steps
Production workflows - When flows are used for important business processes
Consider Disabling Validation When:
Partial results are acceptable - When you'd rather get some output than none at all
Testing and debugging - When you want to see what a step produces even if it's not perfect
Non-critical steps - When a step's failure shouldn't stop the entire flow
Iterative development - When you're still refining your flow and expect some failures
How to Configure Result Validation
Enabling/Disabling Validation
Open your flow in the flow builder
Click on the step you want to configure
Look for the validation setting in the step configuration panel
Toggle the "Result Validation" checkbox to enable or disable it
Best Practices for Validation Configuration
Start with validation enabled for all steps during initial development
Test your flow thoroughly with validation on to identify potential issues
Selectively disable validation only for steps where partial results are acceptable
Document your validation decisions so team members understand why certain steps have validation disabled
Troubleshooting Validation Issues
When Validation Fails
If your flow stops due to validation failure:
Check the step's output - Look at what the step actually produced
Review the "Thinking" tab - Examine the AI's reasoning process
Analyze the instructions - Ensure your step instructions are clear and achievable
Consider the input data - Verify that the step received appropriate input
Refine your prompts - Adjust instructions to be more specific or handle edge cases
Common Validation Failure Scenarios
Incomplete outputs - The step didn't finish generating the expected content
Format issues - The output doesn't match the expected structure or format
Quality concerns - The content quality doesn't meet the validator's standards
Context problems - The step couldn't properly process the input from previous steps
Debugging Strategies
Test individual steps - Run problematic steps in isolation to identify issues
Simplify instructions - Break complex tasks into smaller, more manageable steps
Add explicit formatting requirements - Specify exactly what format you expect
Provide examples - Include sample outputs in your step instructions
Temporarily disable validation - Use this to see partial outputs and understand what's happening
Impact on Performance and Costs
Token Usage
Result validation does use additional tokens as it requires the AI to analyze and evaluate each step's output. However, this cost is typically offset by:
Preventing wasted processing on subsequent steps with bad input
Reducing manual review time by catching errors early
Improving overall flow reliability and reducing rework
Processing Time
Validation adds a small amount of processing time to each step, but this is usually minimal compared to the time saved by preventing flow failures and ensuring quality outputs.
By understanding and properly configuring result validation, you can build more reliable flows that consistently produce high-quality outputs while giving you the flexibility to handle edge cases and partial results when needed.