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_gpu_similarity.py

@@ -0,0 +1,95 @@
+"""Tests for fusionagi.gpu.tensor_similarity."""
+
+import pytest
+
+from fusionagi.gpu.backend import reset_backend, get_backend
+from fusionagi.gpu.tensor_similarity import (
+    pairwise_text_similarity,
+    deduplicate_claims,
+    nearest_neighbors,
+)
+
+
+@pytest.fixture(autouse=True)
+def _use_numpy():
+    reset_backend()
+    get_backend(force="numpy")
+    yield
+    reset_backend()
+
+
+class TestPairwiseTextSimilarity:
+    def test_basic(self):
+        sim = pairwise_text_similarity(["hello world"], ["hello world"])
+        assert sim.shape == (1, 1)
+        assert sim[0, 0] > 0.9
+
+    def test_different_texts(self):
+        sim = pairwise_text_similarity(["hello world"], ["completely different text"])
+        assert sim.shape == (1, 1)
+        assert sim[0, 0] < 1.0
+
+    def test_multi(self):
+        sim = pairwise_text_similarity(
+            ["cat", "dog"],
+            ["car", "bike", "train"],
+        )
+        assert sim.shape == (2, 3)
+
+
+class TestDeduplicateClaims:
+    def test_empty(self):
+        assert deduplicate_claims([]) == []
+
+    def test_single(self):
+        groups = deduplicate_claims(["one claim"])
+        assert groups == [[0]]
+
+    def test_identical(self):
+        groups = deduplicate_claims(
+            ["the sky is blue", "the sky is blue"],
+            threshold=0.9,
+        )
+        assert len(groups) == 1
+        assert sorted(groups[0]) == [0, 1]
+
+    def test_different(self):
+        groups = deduplicate_claims(
+            ["the sky is blue", "python is a programming language"],
+            threshold=0.99,
+        )
+        assert len(groups) == 2
+
+    def test_all_indices_covered(self):
+        claims = ["a", "b", "c", "d"]
+        groups = deduplicate_claims(claims, threshold=0.99)
+        all_indices = sorted(idx for group in groups for idx in group)
+        assert all_indices == [0, 1, 2, 3]
+
+
+class TestNearestNeighbors:
+    def test_empty_query(self):
+        result = nearest_neighbors([], ["corpus text"])
+        assert result == []
+
+    def test_empty_corpus(self):
+        result = nearest_neighbors(["query"], [])
+        assert result == [[]]
+
+    def test_basic(self):
+        result = nearest_neighbors(
+            ["hello world"],
+            ["hello world", "goodbye moon", "hello planet"],
+            top_k=2,
+        )
+        assert len(result) == 1
+        assert len(result[0]) == 2
+        # Each result is (index, score)
+        assert isinstance(result[0][0], tuple)
+        assert isinstance(result[0][0][0], int)
+        assert isinstance(result[0][0][1], float)
+
+    def test_top_k_limit(self):
+        corpus = [f"text {i}" for i in range(20)]
+        result = nearest_neighbors(["text 5"], corpus, top_k=3)
+        assert len(result[0]) == 3