Add quick start guide and architecture overview

README.md

@@ -1,4 +1,7 @@
 # flowforge
+
+See the [quick start](docs/getting-started-quick.md) and [architecture overview](docs/diagrams/overview.svg).
+
 ## 🎯 **The brutal truth about Data Engineering today**
 **Data engineering is broken.** And we're all pretending it's fine.
 

docs/getting-started-quick.md

@@ -0,0 +1,22 @@
+# Quick Start
+
+The smallest possible pipeline in FlowForge.
+
+```scala
+import cats.effect.IO
+import com.flowforge.core.pipeline._
+
+object QuickStart extends IOApp.Simple {
+  def run: IO[Unit] =
+    DataPipelineFactory[IO]
+      .source(blob"gs://raw/sales.csv")
+      .contract(SalesContract.strict)
+      .transform(_.filter(_.amount >= 0))
+      .quality(nonNull("id") and unique("id"))
+      .sink(BigQuerySink("analytics.sales"))
+      .build
+      .run
+}
+```
+
+See the [architecture overview](diagrams/overview.svg) for how components fit together.

docs/plan/v100-plan.md

@@ -213,7 +213,7 @@ That’s it. If you execute those nine steps, the tag is legit.
 
 **P1 — DX & docs**
 
-* [ ] Quickstart with 4 commands + expected outputs/screenshots.
+* [ ] Quickstart with 4 commands + expected outputs/screenshots. See [quick guide](../getting-started-quick.md) and [architecture overview](../diagrams/overview.svg).
 * [ ] `mdoc` wired to `examples/` so doc snippets compile in CI.
 * [ ] Compatibility matrix (Spark/Delta/Deequ) and limitations (no UNIQUE). ([Delta Lake][4])
 * [ ] Comparison page vs Frameless/Dagster/Kedro/Scio—what FlowForge uniquely provides (contracts → enforcement ×3). ([GitHub][9], [Dagster Docs][11], [Kedro][13], [Spotify Open Source][15])