Add full online Phase1 Phase2 final graph extraction pipeline

tests/test_gwam_realtime_env_graph.py

@@ -7,19 +7,47 @@ import numpy as np
 
 sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "realtime"))
 
-from gwam_realtime_env_graph import encode_rle, to_gnn_graph, visibility_change_edges
+from gwam_realtime_env_graph import (
+    build_online_visual_features,
+    encode_rle,
+    to_gnn_graph,
+    visibility_change_edges,
+)
 
 
-def test_encode_rle_starts_with_zero_run_and_roundtrips_simple_mask():
-    mask = np.array([[False, True, True], [False, False, True]], dtype=bool)
-    rle = encode_rle(mask)
-    assert rle == [1, 2, 2, 1]
+class FakeVisualBackend:
+    def image_embedding(self, rgb_frame):
+        return np.asarray(rgb_frame, dtype=np.float32)
+
+    def pool_mask(self, image_embedding, mask):
+        if not mask.any():
+            return None
+        seed = float(image_embedding[mask].mean()) if image_embedding.ndim == 3 else float(mask.mean())
+        vec = np.linspace(0.0, 1.0, 256, dtype=np.float32) + seed / 255.0
+        vec /= np.linalg.norm(vec).clip(min=1e-6)
+        return vec.astype(np.float16)
+
+    def type_clip32(self, nodes):
+        arr = np.zeros((256, 32), dtype=np.float32)
+        for i in range(min(len(nodes), 256)):
+            arr[i, i % 32] = 1.0
+        return arr, {"model": "fake", "projection_seed": 20260702}
+
+
+def _decode_for_test(rle, shape):
     vals = []
     cur = 0
     for count in rle:
-        vals.extend([cur] * count)
+        vals.extend([cur] * int(count))
         cur = 1 - cur
-    assert np.array(vals, dtype=bool).reshape(mask.shape).tolist() == mask.tolist()
+    return np.array(vals, dtype=bool).reshape(shape)
+
+
+def test_encode_rle_starts_with_zero_run_and_roundtrips_simple_mask():
+    mask = np.array([[False, True, True], [False, False, True]], dtype=bool)
+    rle = encode_rle(mask)
+    assert rle == [1, 2, 2, 1]
+    assert _decode_for_test(rle, mask.shape).tolist() == mask.tolist()
 
 
 def test_visibility_change_edges_are_self_edges_by_view():
@@ -51,3 +79,78 @@ def test_to_gnn_graph_active_masks_family_onehot_and_edges():
     assert graph["edge_attr"].shape == (2, 8)
     assert "family_idx" not in graph["feature_schema"]
     assert "family_onehot_9" in graph["feature_schema"]
+
+
+def test_online_visual_features_sparse_block_matches_phase2_policy_with_fake_backend():
+    mask = np.zeros((256, 256), dtype=bool)
+    mask[20:40, 30:50] = True
+    rgb_frames = {0: np.full((256, 256, 3), 42, dtype=np.uint8)}
+    rle_rows = [{"n": 1, "v": 0, "rle": encode_rle(mask)}]
+    nodes = [{"category_text": "robot"}, {"category_text": "drawer handle"}]
+    visual = build_online_visual_features(
+        rgb_frames=rgb_frames,
+        rle_rows=rle_rows,
+        nodes=nodes,
+        visual_backend=FakeVisualBackend(),
+        t=0,
+    )
+    assert visual["t"].tolist() == [0]
+    assert visual["n"].tolist() == [1]
+    assert visual["v"].tolist() == [0]
+    assert visual["feat"].shape == (1, 256)
+    assert visual["feat"].dtype == np.float16
+    assert visual["type_clip32"].shape == (256, 32)
+    assert visual["visual_computed"].tolist() == [True]
+    assert visual["summary"]["features_written"] == 1
+    assert visual["summary"]["invalid_visible_pairs"] == 0
+    assert visual["summary"]["invisible_features_injected"] == 0
+
+
+def test_final_fused_graph_has_packaged_loader_342d_feature_layout():
+    node_state = np.zeros((5, 32), dtype=np.float32)
+    node_state[[0, 1], 0] = 1
+    node_state[[0, 1], 1] = [9, 6]
+    node_state[[0, 1], 2:22] = np.arange(40, dtype=np.float32).reshape(2, 20)
+
+    view_visible = np.zeros((5, 3), dtype=bool)
+    view_visible[1, 0] = True
+    view_centroid = np.zeros((5, 3, 2), dtype=np.float16)
+    view_area = np.zeros((5, 3, 1), dtype=np.float16)
+    view_bbox = np.zeros((5, 3, 4), dtype=np.float16)
+    view_centroid[1, 0] = [0.25, 0.5]
+    view_area[1, 0, 0] = 0.1
+    view_bbox[1, 0] = [0.1, 0.2, 0.3, 0.4]
+
+    mask = np.zeros((256, 256), dtype=bool)
+    mask[10:20, 10:20] = True
+    visual = build_online_visual_features(
+        rgb_frames={0: np.full((256, 256, 3), 7, dtype=np.uint8)},
+        rle_rows=[{"n": 1, "v": 0, "rle": encode_rle(mask)}],
+        nodes=[{"category_text": "robot"}, {"category_text": "handle"}],
+        visual_backend=FakeVisualBackend(),
+        t=0,
+    )
+    graph = to_gnn_graph(
+        node_state=node_state,
+        static_edges=[{"src": 1, "dst": 0, "rel": "part_of"}],
+        dynamic_edges=[{"src": 0, "dst": 1, "rel": "contact"}],
+        view_visible=view_visible,
+        view_centroid=view_centroid,
+        view_area=view_area,
+        view_bbox=view_bbox,
+        visual_features=visual,
+        include_visual=True,
+        include_type_clip32=True,
+        include_full_view_evidence=True,
+    )
+    assert graph["slot_ids"].tolist() == [0, 1]
+    assert graph["x"].shape == (2, 342)
+    assert len(graph["feature_schema"]) == 342
+    assert graph["visual_feature_valid"].tolist() == [False, True]
+    assert graph["metadata"]["include_visual"] is True
+    assert graph["metadata"]["include_type_clip32"] is True
+    assert graph["metadata"]["include_full_view_evidence"] is True
+    assert graph["edge_attr"].shape == (2, 8)
+    assert graph["feature_schema"][30] == "visual_sam2_mean_000"
+    assert graph["feature_schema"][286] == "type_clip32_000"
+    assert graph["feature_schema"][318] == "view_right_visible"