Unwind
Decompose array fields into separate documents for per-element processing
On one real run
Three Results Went In. Seven Rows Came Out: What Unwind Actually Does to a Result Set
One search for a nylon hiking backpack returns three documents. Two carry no edges and are dropped entirely by the unwind stage; the third carries seven edges and becomes seven rows. input_count 3, output_count 7, and the two dropped documents never appear as empty rows because preserve_null_and_empty defaults to false.
See the full diagram
Configuration
REQUIRED. Array field to unwind.
Keep documents where the field is null or missing.
Add an index field tracking the element's position.
Output
No output yet
Configure the stage parameters and click "Run Stage" to see the simulated output.
Related to unwind
Pages on the same stages, extractors and patterns.
Json Transform
Transform documents using Jinja2 templates that render to JSON
Rag Prepare
Prepare documents for LLM context windows with token management
External Web Search
Enrich documents with web search results using Exa AI
Api Call
Enrich documents with external API calls
Sql Lookup
Enrich documents with SQL query results from external databases
Llm Enrich
Enrich documents with LLM-generated fields using natural language prompts