The Array Flatten node collects results from Array Splitter processing, combining individual items back into a single array output. This is essential for completing batch processing workflows.

How It Works

Array Flatten is the counterpart to Array Splitter:

Array Splitter:
  Input:  { items: [1, 2, 3] }
  Output: Three separate: 1, 2, 3

Processing: Each item processed individually

Array Flatten:
  Input:  Three separate: "result1", "result2", "result3"
  Output: { results: ["result1", "result2", "result3"] }

Configuration

Array Flatten automatically detects and collects outputs from upstream Array Splitter operations. No manual configuration is required.

Output Structure

{
  "out": [
    "first processed item",
    "second processed item",
    "third processed item"
  ],
  "count": 3,
  "originalInput": {...}
}

Common Usage Patterns

Complete Processing Pipeline

Dataset Source (items)
  → Array Splitter (split items)
  → Prompt (process each)
  → Array Flatten (collect results)
  → Dataset Sink (save all)

Multi-Step Processing

Website Mapper (discover URLs)
  → Array Splitter (split URLs)
  → Page Scraper (scrape each)
  → HTML Text Extractor (clean each)
  → Prompt (analyze each)
  → Array Flatten (collect analyses)
  → Prompt (summarize all)
  → Dataset Sink

Parallel Branch Collection

Array Splitter
  → Branch A: Prompt 1
  → Branch B: Prompt 2
  → Array Flatten (collects from both branches)

Order Preservation

Array Flatten maintains the order of items from the original Array Splitter:

• Sequential processing: Order is always preserved
• Parallel processing: Results are reordered to match input order
• Failed items: Can be included or excluded based on configuration

Error Handling

Partial Failures

If some items fail during processing:

Input: 5 items split
Process: Items 1, 2, 4, 5 succeed; item 3 fails
Output: 
  - Success array: [result1, result2, result4, result5]
  - Failed items logged separately
  - Total count includes both success and failures

All Items Fail

If all split items fail to process:

• Array Flatten outputs an empty array
• Downstream nodes receive empty input
• Flow continues (doesn't stop)
• Error details available in logs

Performance Considerations

Memory Usage

Array Flatten collects all results in memory:

• Small arrays (<100 items): No issues
• Medium arrays (100-1000 items): Monitor memory
• Large arrays (1000+ items): Consider streaming or batching
• Very large results: May need to write to dataset incrementally

Wait Time

Array Flatten waits for all split items to complete:

• Sequential: Waits for each item in order
• Parallel: Waits for slowest item
• Timeout handling: Configure max wait time if needed

Best Practices

Always Use After Array Splitter

Don't leave Array Splitter outputs uncollected:

✅ Good:
Array Splitter → Process → Array Flatten → Continue

❌ Bad:
Array Splitter → Process → (no flatten) → Dataset Sink
(Only saves last item!)

Place Strategically

Position Array Flatten where you need the full collection:

Option 1: Flatten immediately
Array Splitter → Process → Array Flatten → More Processing

Option 2: Flatten after multiple steps
Array Splitter → Step 1 → Step 2 → Step 3 → Array Flatten

Handle Empty Results

Prepare downstream nodes for potentially empty arrays:

Array Flatten → Check if empty → Continue or skip

Advanced Patterns

Nested Flattening

When using multiple Array Splitters, use matching Array Flattens:

Array Splitter 1 (outer)
  → Array Splitter 2 (inner)
    → Process
    → Array Flatten 2 (collect inner)
  → Array Flatten 1 (collect outer)

Conditional Flattening

Filter or transform items before flattening:

Array Splitter
  → Process
  → Filter (keep only successful)
  → Array Flatten (collect filtered)

Aggregation After Flattening

Use the flattened array for aggregate operations:

Array Flatten (collect all results)
  → Prompt: "Summarize these {count} results: {results}"
  → Generate aggregate report

Debugging Tips

Check Array Sizes

Monitor the count field to verify all items were collected:

Expected: 10 items split
Flattened count: 8 items
→ 2 items failed or were filtered

Inspect Individual Results

Review the flattened array to identify issues:

• Look for null or undefined values
• Check for error messages in results
• Verify data structure consistency
• Compare with expected outputs

Test with Small Arrays

Always test your split-process-flatten logic with 2-3 items first:

• Easier to debug
• Faster iteration
• Clear visibility into each step
• Validates logic before scaling

Integration Examples

With Dataset Sink

Array Flatten
  → Map: out (array) → Dataset Sink
  → Each array item becomes a dataset row

With Prompt for Summary

Array Flatten
  → Prompt: "Analyze these {count} results and provide insights"
  → Generate executive summary

With Further Processing

Array Flatten (collect processed items)
  → Transform or filter array
  → Split again for second-stage processing
  → Flatten again for final output