feat(runtime): capture step evidence in console

runtime/task.py

@@ -136,13 +136,22 @@ class TaskRunner:
                 if should_stop is not None and should_stop():
                     return self._interrupt_task(task)
                 executable_step = self._step_for_device(step, task.device_id)
+                before_screenshot = self._planning_screenshot(task.device_id)
                 result = self.executor.execute(
                     executable_step,
                     context=context,
                 )
+                after_screenshot = self._planning_screenshot(task.device_id)
                 context.add_step_result(result)
                 self._record_step_result(
-                    world_handle, context, task, scene, step, result
+                    world_handle,
+                    context,
+                    task,
+                    scene,
+                    step,
+                    result,
+                    before_screenshot=before_screenshot,
+                    after_screenshot=after_screenshot,
                 )
                 self._emit_step_progress(
                     len(context.step_results),
@@ -204,6 +213,9 @@ class TaskRunner:
         scene: Scene,
         step: PlannedStep,
         result: object,
+        *,
+        before_screenshot: bytes | None = None,
+        after_screenshot: bytes | None = None,
     ) -> None:
         """Record one executed step's bookkeeping: world-model observe + timeline append.
 
@@ -212,7 +224,14 @@ class TaskRunner:
         cannot drift apart on what gets recorded for a step.
         """
         self._update_world(world_handle, context, scene, step, result)
-        self._append_timeline(task, scene, step, result)
+        self._append_timeline(
+            task,
+            scene,
+            step,
+            result,
+            before_screenshot=before_screenshot,
+            after_screenshot=after_screenshot,
+        )
 
     def _emit_step_progress(
         self, step_index: int, step_status: str, summary: str
@@ -330,10 +349,19 @@ class TaskRunner:
         scene: Scene,
         step: PlannedStep,
         result: object,
+        *,
+        before_screenshot: bytes | None = None,
+        after_screenshot: bytes | None = None,
     ) -> None:
         if not self.timeline:
             return
-        screenshot = self._planning_screenshot(task.device_id)
+        ocr_results = scene.ocr_results_to_dict()
+        if not ocr_results:
+            ocr_results = [
+                element.to_dict()
+                for element in scene.elements
+                if element.source == "ocr"
+            ]
         self.timeline.append(
             task_id=task.id,
             scene=scene.to_dict(),
@@ -346,7 +374,9 @@ class TaskRunner:
             result=result.to_dict()
             if hasattr(result, "to_dict")
             else {"result": result},
-            screenshot=screenshot,
+            before_screenshot=before_screenshot,
+            after_screenshot=after_screenshot,
+            ocr_results=ocr_results,
         )
 
     def _step_for_device(self, step: PlannedStep, device_id: str) -> PlannedStep: