Add overnight flux2klein smoke benchmark runner

benchmarks/flux2klein/README.md

@@ -0,0 +1,38 @@
+# FLUX.2-klein Benchmarks
+
+This folder holds durable benchmark inputs, outputs, and state for low-cost `flux2klein` testing.
+
+## Current intent
+
+- Keep overnight tests cheap and reviewable.
+- Prefer one or two single-image smoke tests over broad matrices.
+- Log enough context to reproduce a result without trawling temporary folders.
+
+## Layout
+
+- `canonical-inputs/`
+  - durable reference inputs used for repeatable comparisons
+- `runs/`
+  - dated run batches
+- `overnight-state.json`
+  - handoff file between the `3am` optimisation pass and the `6am` review pass
+
+## Batch contents
+
+Each case folder should keep:
+
+- `input-1.*`
+- `resolved-prompt.json`
+- `run-record.json`
+- the saved output image
+
+## Current nightly default
+
+- family: `flux2klein`
+- workflow: `models/flux2klein/image-edit-9b-distilled/workflow.json`
+- output policy: single image only
+- test style: smoke test, not broad sweep
+
+## Review rule
+
+Do not call a workflow fully optimal from one benchmark image. Use these batches to decide whether a path is promising enough for follow-up.

benchmarks/flux2klein/canonical-inputs/README.md

@@ -0,0 +1,5 @@
+# Canonical Inputs
+
+Put durable benchmark reference images here when a test input becomes part of the standing comparison set.
+
+For now, the overnight runner can also copy an external source image into each run case so the used input is still captured in-repo even before it graduates into this folder.

benchmarks/flux2klein/overnight-state.json

@@ -0,0 +1,10 @@
+{
+  "date": null,
+  "testsSet": false,
+  "serverBusy": false,
+  "reason": "No overnight run has written state yet.",
+  "cases": [],
+  "reviewPending": false,
+  "optimizedPaths": [],
+  "verdict": null
+}