Let effects declare whether they need source images

Whether an effect works from the loaded photographs is now part of the
declarative plug-in contract, next to params(): Effect::needsSources()
defaults to true, and blur — which only softens what it is handed — says
no and blurs an empty canvas when given nothing.

Only the first stage of a cascade ever sees the sources, so
chainNeedsSources() just asks the first stage. Both front-ends gate on
that instead of on "are there any images": the CLI only refuses an empty
batch when the chain wants one, and the interface renders a blur-led
chain with no images loaded, saying whichever thing is actually missing.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

src/effects/blur.cpp

@@ -100,7 +100,6 @@ void blurVertical(const Image& src, Image& dst, int radius) {
 
 Image BlurEffect::render(const std::vector<Image>& sources, int width, int height, Rng& rng,
                          const ParamValues& values) const {
-    if (sources.empty()) throw EffectError("No source images.");
     if (width < 1 || height < 1) throw EffectError("Canvas size must be at least 1x1 pixels.");
 
     const int radius = static_cast<int>(std::max(1L, paramInt(values, "radius", 8)));
@@ -108,15 +107,16 @@ Image BlurEffect::render(const std::vector<Image>& sources, int width, int heigh
     const std::string direction = paramChoice(values, "direction", "both");
     const bool firstSource = paramBool(values, "first_source", false);
 
-    // Every random value is drawn here, before any threading, so that a seed
-    // always gives the same image.
-    const std::size_t index = firstSource ? 0 : rng.below(sources.size() - 1);
-    const Image& src = sources[index];
-    const int sx = static_cast<int>(rng.inRange(0, std::max(0, src.width() - width)));
-    const int sy = static_cast<int>(rng.inRange(0, std::max(0, src.height() - height)));
-
     Image canvas(width, height);
-    fillFromSource(canvas, src, sx, sy);
+    if (!sources.empty()) {
+        // Every random value is drawn here, before any threading, so that a
+        // seed always gives the same image.
+        const std::size_t index = firstSource ? 0 : rng.below(sources.size() - 1);
+        const Image& src = sources[index];
+        const int sx = static_cast<int>(rng.inRange(0, std::max(0, src.width() - width)));
+        const int sy = static_cast<int>(rng.inRange(0, std::max(0, src.height() - height)));
+        fillFromSource(canvas, src, sx, sy);
+    }
 
     Image scratch(width, height);
     for (int pass = 0; pass < passes; ++pass) {

src/effects/blur.hpp

@@ -4,17 +4,19 @@
 
 namespace effector {
 
-// Fills the canvas with one of the sources and softens it with a separable box
-// blur, repeated a few times to approximate a gaussian.
+// Softens its input with a separable box blur, repeated a few times to
+// approximate a gaussian. It needs no source images of its own: given none it
+// blurs an empty canvas, which is what makes it usable as a pure filter stage.
 class BlurEffect : public Effect {
 public:
     std::string name() const override { return "blur"; }
     std::string title() const override { return "Blur"; }
     std::string description() const override {
-        return "Fills the canvas from a randomly chosen source and softens it with a repeated "
-               "box blur, horizontally, vertically or both.";
+        return "Softens its input with a repeated box blur, horizontally, vertically or both; "
+               "with sources loaded it fills the canvas from a randomly chosen one first.";
     }
     const std::vector<ParamSpec>& params() const override;
+    bool needsSources() const override { return false; }
 
     Image render(const std::vector<Image>& sources, int width, int height, Rng& rng,
                  const ParamValues& values) const override;