An AutoTable pipeline is a named MongoDB aggregation definition stored on a schema container and executed against that project-scoped collection to produce read-only, shaped result documents.
What a pipeline does
A pipeline receives every source document as input to its first stage. Each stage filters, calculates, groups, joins, sorts, or reshapes the stream, and its output becomes the next stage's input. The final documents are returned as a JSON array.
Use pipelines for reports, dashboards, enriched lists, rollups, and page/table data sources. A pipeline is not a workflow: it does not react to lifecycle triggers, mutate business state through ordered steps, or provide transactional/outbox execution modes.
Stored pipeline
A named PipelineStage on a schema container, callable through the dynamic pipeline endpoint and eligible for pipeline caching.
Workflow aggregate
An aggregate/run_pipeline workflow step with workflow templates, an allowed-stage check, and an automatic 500-result safety limit.
Schema query
A normal filtered collection read with row access, search, sort, and pagination rather than an aggregation stage chain.
text
source collection
→ $match eligible documents
→ $group business totals
→ $project public result fields
→ $sort deterministic order
→ result document array
Core contract
Execution requires tenant, project, schemaName, and pipelineName. The service finds the named PipelineStage, prepares placeholders, optionally serves or fills a cache entry, executes Mongo aggregation, and returns result objects. Invalid JSON, an unsupported Mongo expression, a missing collection, or a database error fails the request.
Read the result shape, not the source schema
Aggregation stages may remove, rename, nest, or synthesize fields. Consumers should bind to the final output contract rather than assuming source fields survive.