Once you've built your flow, running it executes all nodes in sequence, processing your data through each step while providing real-time progress feedback.

How to Run a Flow

1. 1

   Select data source If using Dataset Source, select which samples to process

2. 2

   Click "Run Flow" The execution button in the top toolbar

3. 3

   Monitor progress Watch real-time status updates for each node

4. 4

   Review results Check outputs in Dataset Sinks or node inspectors

Execution Modes

Sync (Synchronous) Mode

Default mode for manual flow execution:

• Waits for completion before returning
• Shows progress in real-time
• Best for testing and development
• Limited to reasonable execution times (<5 minutes)

Async (Asynchronous) Mode

Used for long-running flows and API deployments:

• Returns immediately with execution ID
• Flow continues in background
• Poll for status updates
• Best for large datasets or slow operations

Progress Monitoring

Node Status Indicators

Each node shows its current status during execution:

• Pending: Waiting to execute
• Running: Currently processing (animated)
• Success: Completed successfully (green checkmark)
• Error: Failed with error (red X)
• Skipped: Not executed due to conditions

Progress Metrics

Monitor key execution metrics:

Real-Time Logs

View detailed logs during execution:

• Node start/completion times
• Data flow between nodes
• Error messages and warnings
• Resource usage (tokens, API calls)

Running with Datasets

Sample Selection

Choose which dataset rows to process:

Options:
  - Selected samples (manually picked)
  - First N samples (for testing)
  - Random sample (statistical testing)
  - All samples (full dataset run)

Incremental Processing

Process dataset in batches:

1. Run with 5 samples (validate logic)
2. Run with 20 samples (test edge cases)
3. Run with 100 samples (verify scale)
4. Run full dataset (production)

Execution Control

Pause/Resume

For long-running flows:

• Pause execution at any time
• Inspect intermediate results
• Resume from where you paused
• Useful for debugging complex flows

Stop/Cancel

Abort execution if needed:

• Stops current node immediately
• Saves partial results
• Can restart from beginning
• Refunds unused credits

Retry Failed Nodes

Recover from transient failures:

Error on Node 5 (API timeout)
  → Click "Retry Node"
  → Continues from failure point
  → Doesn't re-execute successful nodes

Performance Optimization

Parallel vs Sequential

Sequential (default):
  Process items one at a time
  Lower credit usage
  Easier to debug
  Time: N × per-item-time

Parallel:
  Process multiple items simultaneously  
  Higher credit usage
  Faster completion
  Time: ceil(N / concurrency) × per-item-time

Caching

Evaligo caches certain node outputs:

• Website mapper results (24 hours)
• Page scraper content (1 hour)
• Prompt responses (optional, configurable)
• Reduces redundant API calls

Resource Limits

Be aware of execution limits:

Free Tier:
  - Max execution time: 5 minutes
  - Max concurrent runs: 2
  - Max items per array: 100

Pro Tier:
  - Max execution time: 30 minutes
  - Max concurrent runs: 10
  - Max items per array: 1,000

Enterprise:
  - Custom limits
  - Dedicated resources

Result Inspection

Node Outputs

Click any node after execution to view:

• Input data received
• Output data produced
• Execution time
• Resource usage (tokens, cost)

Dataset Sink Results

View saved results in datasets:

After flow completes:
  → Navigate to Datasets
  → Open target dataset
  → See newly added rows
  → Export or further analyze

Error Analysis

When nodes fail:

• Click failed node for error details
• Check error type and message
• View input data that caused failure
• Fix and retry

Execution History

Access past flow runs:

• View all executions with timestamps
• Compare results across runs
• Replay previous executions
• Export execution data

Best Practices

Pre-Flight Checks

Before running a flow:

1. Verify all node configurations
2. Check variable mappings are correct
3. Ensure datasets have valid data
4. Review prompt configurations
5. Check API keys are active

Test Incrementally

Step 1: Test with 1 sample
Step 2: Test with 5 samples  
Step 3: Test with different data types
Step 4: Test edge cases
Step 5: Run full dataset

Monitor Costs

Track API credit usage:

• Estimate cost before running
• Monitor usage during execution
• Set up cost alerts
• Review cost per execution