Unify harness L-C-A on the canonical lca.WorkloadProfile

Phase 0 of the two-stop work. The prompt block labeled `workload_lca_profile`
previously re-derived L-C-A from summarize_window's ad-hoc percentiles, diverging
from the paper's 10-dim RobustScaler vector implemented in lca.py. Make that block
authoritative: build_harness_context now accepts an optional workload_profile and
renders the canonical 10-dim vector + per-family stats when present, falling back
to the legacy rendering only when no profile is supplied (direct unit-test calls).

Real call sites (study prompt/llm-propose/tune, run_baseline_then_llm) build the
profile via lca.build_study_workload_profile and pass it through build_prompt. The
heuristic regime classifiers keep reading window_summary; that is the heuristic
layer, distinct from the similarity metric.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

tests/test_core_flow.py

@@ -28,6 +28,7 @@ from aituner.harness import (
     build_harness_stop_proposal,
 )
 from aituner.lca import (
+    build_study_workload_profile,
     build_workload_profile,
     profile_similarity,
     resolve_length_mode,
@@ -298,6 +299,34 @@ class CoreFlowTests(unittest.TestCase):
         self.assertAlmostEqual(profile.stats["arrival"]["fano_1s"], 0.5)
         self.assertEqual(resolve_length_mode(request_mode="decode_only"), "output")
 
+    def test_harness_context_uses_canonical_lca_vector(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            tmp_path = Path(tmp)
+            study_path = _write_study_assets(tmp_path)
+            study = load_study_spec(study_path)
+            window, requests = load_trace_requests(study, study_spec_path=study_path)
+            profile = build_study_workload_profile(study, requests, window)
+            state = StudyState(study_id=study.study_id, trials=[])
+            summary = summarize_window(requests, window)
+            context = build_harness_context(
+                study=study,
+                window_summary=summary,
+                state=state,
+                workload_profile=profile,
+            )
+            block = context["workload_lca_profile"]
+            # The labeled L-C-A block is the canonical 10-dim metric, not ad-hoc.
+            self.assertEqual(block["vector"], profile.vector)
+            self.assertEqual(len(block["vector"]), 10)
+            self.assertIn("RobustScaler", block["metric"])
+            # Without a profile it falls back to the legacy ad-hoc rendering.
+            legacy = build_harness_context(
+                study=study,
+                window_summary=summary,
+                state=state,
+            )["workload_lca_profile"]
+            self.assertNotIn("vector", legacy)
+
     def test_lca_similarity_matrix_separates_different_profiles(self) -> None:
         window = WindowRecord(
             window_id="base",