feat: GPU/TensorCore integration — TensorFlow backend, GPU-accelerated reasoning, training, and memory

- New fusionagi/gpu/ module with TensorBackend protocol abstraction
  - TensorFlowBackend: GPU-accelerated ops with TensorCore mixed-precision
  - NumPyBackend: CPU fallback (always available, no extra deps)
  - Auto-selects best available backend at runtime

- GPU-accelerated operations:
  - Cosine similarity matrix (batched, XLA-compiled)
  - Multi-head attention for consensus scoring
  - Batch hypothesis scoring on GPU
  - Semantic similarity search (pairwise, nearest-neighbor, deduplication)

- New TensorFlowAdapter (fusionagi/adapters/):
  - LLMAdapter for local TF/Keras model inference
  - TensorCore mixed-precision support
  - GPU-accelerated embedding synthesis fallback

- Reasoning pipeline integration:
  - gpu_scoring.py: drop-in GPU replacement for multi_path scoring
  - Super Big Brain: use_gpu config flag, GPU scoring when available

- Memory integration:
  - gpu_search.py: GPU-accelerated semantic search for SemanticGraphMemory

- Self-improvement integration:
  - gpu_training.py: gradient-based heuristic weight optimization
  - Reflective memory training loop with loss tracking

- Dependencies: gpu extra (tensorflow>=2.16, numpy>=1.26)
- 64 new tests (276 total), all passing
- Architecture spec: docs/gpu_tensorcore_integration.md

Co-Authored-By: Nakamoto, S <defi@defi-oracle.io>

tests/test_tensorflow_adapter.py

@@ -0,0 +1,77 @@
+"""Tests for fusionagi.adapters.tensorflow_adapter (uses NumPy backend, no TF required)."""
+
+import pytest
+
+from fusionagi.gpu.backend import reset_backend, get_backend
+
+
+@pytest.fixture(autouse=True)
+def _use_numpy():
+    reset_backend()
+    get_backend(force="numpy")
+    yield
+    reset_backend()
+
+
+class TestTensorFlowAdapterImport:
+    """Test that TensorFlowAdapter is importable (may be None without TF)."""
+
+    def test_import(self):
+        from fusionagi.adapters import TensorFlowAdapter
+        # TensorFlowAdapter is None when tensorflow is not installed
+        # This is by design — GPU is an optional dependency
+
+
+class TestGPUMemorySearch:
+    """Test GPU-accelerated memory search."""
+
+    def test_semantic_search(self):
+        from fusionagi.memory.gpu_search import semantic_search
+        from fusionagi.schemas.atomic import AtomicSemanticUnit, AtomicUnitType
+
+        units = [
+            AtomicSemanticUnit(
+                unit_id="u1",
+                content="the sky is blue",
+                type=AtomicUnitType.FACT,
+                confidence=1.0,
+            ),
+            AtomicSemanticUnit(
+                unit_id="u2",
+                content="water is wet",
+                type=AtomicUnitType.FACT,
+                confidence=1.0,
+            ),
+            AtomicSemanticUnit(
+                unit_id="u3",
+                content="python programming language",
+                type=AtomicUnitType.FACT,
+                confidence=1.0,
+            ),
+        ]
+        results = semantic_search("sky color", units, top_k=2)
+        assert len(results) <= 2
+        assert all(isinstance(r, tuple) for r in results)
+        assert all(isinstance(r[0], AtomicSemanticUnit) for r in results)
+        assert all(isinstance(r[1], float) for r in results)
+
+    def test_semantic_search_empty(self):
+        from fusionagi.memory.gpu_search import semantic_search
+
+        results = semantic_search("query", [], top_k=5)
+        assert results == []
+
+    def test_batch_embed_units(self):
+        from fusionagi.memory.gpu_search import batch_embed_units
+        from fusionagi.schemas.atomic import AtomicSemanticUnit, AtomicUnitType
+
+        units = [
+            AtomicSemanticUnit(
+                unit_id="u1",
+                content="test content",
+                type=AtomicUnitType.FACT,
+                confidence=1.0,
+            ),
+        ]
+        result = batch_embed_units(units)
+        assert result is not None