Pipeline examples¶
Choose a task, then use its canonical recipe to set up, validate, deploy, run and inspect a pipeline with synthetic data.
| Task | Flow | Recipe |
|---|---|---|
| Enrich API orders while keeping unknown customers | HTTP + PostgreSQL → S3 JSON | Customer enrichment |
| Reconcile payments, refunds and unmatched ledger entries | S3 CSV + PostgreSQL → BigQuery | Settlement reconciliation |
| Explore normalization, nullable comparisons, aggregation and branching lineage | S3 CSV + PostgreSQL → two PostgreSQL upserts | Advanced settlement review (BP-1011 demo) |
| Apply spreadsheet overrides before CRM upsert | S3 XLSX + PostgreSQL → HubSpot | Contact overrides |
| Project document inventory into relational reporting | MongoDB → MySQL | Inventory projection |
| Archive a file without changing its bytes | HTTP → S3 | File archive |
The public examples repository owns the complete specifications, operations, binding templates, fixtures, expected results and qualification records. Its manifest is the entry point for automation. These pages explain which recipe to choose; they do not maintain another copy of its spec.
Before you run¶
Use isolated non-production systems and synthetic data. Start with the repository's shared setup and execution guide. It covers authentication, Git authoring, connections, saved operations, logical bindings, target validation, exact-version deployment, run inspection and cleanup. Provider permissions and destination effects are separate from Braidplane permissions.
Use the published public CLI v0.1.0-alpha.6, pinned by the examples repository. Public CLI alpha.5 lacks the required left-join support. See CLI installation and pipeline specifications for the platform fundamentals.
What the evidence establishes¶
The initial five specs passed source-built CLI validation and controlled local execution with independent destination comparisons. The sixth, advanced BP-1011 demo retains its separate historical local rehearsal and independent reporting-table reads; its spec passed fresh public alpha.6 offline validation. PostgreSQL, MongoDB, MySQL and S3-compatible storage used local containers. HTTP used a synthetic fixture; BigQuery and HubSpot used local protocol mocks. These records do not establish acceptance in Google Cloud or a HubSpot account, provider retries, or production readiness.
Each recipe keeps its exact executed version/run identity and complete step evidence. A new run requires its own evidence. Read the coverage and friction ledger, including the observed transform-contract failure and missing-binding rejection, before adapting a recipe.
The earlier PostgreSQL-to-S3 walkthrough remains a separate guided introduction. Its downloadable assets describe that tutorial; they are not alternative copies of this collection.