feat: source semantic hardening — A→B→C staged L2 with domain + few-shot

.env.example

@@ -19,3 +19,10 @@ EMBEDDING_PROVIDER=openrouter
 EMBEDDING_MODEL=google/gemini-embedding-001
 EMBEDDING_API_KEY=sk-or-...
 # EMBEDDING_BASE_URL=  # Only needed for databricks or custom providers
+
+# Ingest / data bridge (optional)
+# INGEST_DUCKDB_PATH=~/.sema/poc.duckdb
+# INGEST_CACHE_DIR=~/.sema/cache/cbioportal
+# INGEST_OMOP_VOCAB_PATH=~/data/omop/athena_2026_04
+# Required only for vocabulary push via COPY INTO (falls back to INSERT when unset).
+# INGEST_CLOUD_STAGING_URI=s3://my-bucket/sema-staging/

.gitignore

@@ -31,6 +31,7 @@ Thumbs.db
 
 # Project tooling
 openspec/
+.wolf/
 .mypy_cache
 .pytest_cache
 htmlcov/
@@ -40,3 +41,6 @@ coverage.xml
 # Docs (generated, source lives elsewhere)
 docs/
 backups/
+
+# Eval run logs (artifacts kept, logs are transient)
+eval-runs/*.log

eval-runs/end-to-end-diff.json

@@ -0,0 +1,72 @@
+{
+  "summary": {
+    "tables_compared": 6,
+    "total_added": 50,
+    "total_removed": 94,
+    "total_changed": 814,
+    "only_in_baseline": [],
+    "only_in_current": []
+  },
+  "per_table": [
+    {
+      "table": "mutation",
+      "summary": {
+        "added_count": 25,
+        "removed_count": 51,
+        "changed_count": 385,
+        "total_before": 800,
+        "total_after": 799
+      }
+    },
+    {
+      "table": "patient",
+      "summary": {
+        "added_count": 12,
+        "removed_count": 16,
+        "changed_count": 142,
+        "total_before": 362,
+        "total_after": 406
+      }
+    },
+    {
+      "table": "sample",
+      "summary": {
+        "added_count": 5,
+        "removed_count": 3,
+        "changed_count": 72,
+        "total_before": 177,
+        "total_after": 213
+      }
+    },
+    {
+      "table": "timeline_sample_acquisition",
+      "summary": {
+        "added_count": 0,
+        "removed_count": 8,
+        "changed_count": 48,
+        "total_before": 122,
+        "total_after": 103
+      }
+    },
+    {
+      "table": "timeline_status",
+      "summary": {
+        "added_count": 2,
+        "removed_count": 9,
+        "changed_count": 59,
+        "total_before": 153,
+        "total_after": 189
+      }
+    },
+    {
+      "table": "timeline_treatment",
+      "summary": {
+        "added_count": 6,
+        "removed_count": 7,
+        "changed_count": 108,
+        "total_before": 233,
+        "total_after": 294
+      }
+    }
+  ]
+}

eval-runs/post-cleanup-diff.json

@@ -0,0 +1,132 @@
+{
+  "summary": {
+    "tables_compared": 12,
+    "total_added": 23,
+    "total_removed": 22,
+    "total_changed": 670,
+    "only_in_baseline": [],
+    "only_in_current": []
+  },
+  "per_table": [
+    {
+      "table": "clinical_supp_hypoxia",
+      "summary": {
+        "added_count": 0,
+        "removed_count": 0,
+        "changed_count": 9,
+        "total_before": 44,
+        "total_after": 44
+      }
+    },
+    {
+      "table": "cna",
+      "summary": {
+        "added_count": 0,
+        "removed_count": 0,
+        "changed_count": 8,
+        "total_before": 42,
+        "total_after": 41
+      }
+    },
+    {
+      "table": "gene_panel_matrix",
+      "summary": {
+        "added_count": 0,
+        "removed_count": 0,
+        "changed_count": 17,
+        "total_before": 44,
+        "total_after": 43
+      }
+    },
+    {
+      "table": "mutation",
+      "summary": {
+        "added_count": 14,
+        "removed_count": 11,
+        "changed_count": 256,
+        "total_before": 788,
+        "total_after": 795
+      }
+    },
+    {
+      "table": "patient",
+      "summary": {
+        "added_count": 1,
+        "removed_count": 0,
+        "changed_count": 116,
+        "total_before": 364,
+        "total_after": 362
+      }
+    },
+    {
+      "table": "resource_definition",
+      "summary": {
+        "added_count": 0,
+        "removed_count": 0,
+        "changed_count": 17,
+        "total_before": 57,
+        "total_after": 58
+      }
+    },
+    {
+      "table": "resource_patient",
+      "summary": {
+        "added_count": 0,
+        "removed_count": 0,
+        "changed_count": 9,
+        "total_before": 31,
+        "total_after": 30
+      }
+    },
+    {
+      "table": "sample",
+      "summary": {
+        "added_count": 0,
+        "removed_count": 1,
+        "changed_count": 50,
+        "total_before": 184,
+        "total_after": 182
+      }
+    },
+    {
+      "table": "structural_variant",
+      "summary": {
+        "added_count": 0,
+        "removed_count": 0,
+        "changed_count": 38,
+        "total_before": 117,
+        "total_after": 115
+      }
+    },
+    {
+      "table": "timeline_sample_acquisition",
+      "summary": {
+        "added_count": 5,
+        "removed_count": 1,
+        "changed_count": 44,
+        "total_before": 118,
+        "total_after": 118
+      }
+    },
+    {
+      "table": "timeline_status",
+      "summary": {
+        "added_count": 0,
+        "removed_count": 5,
+        "changed_count": 25,
+        "total_before": 99,
+        "total_after": 77
+      }
+    },
+    {
+      "table": "timeline_treatment",
+      "summary": {
+        "added_count": 3,
+        "removed_count": 4,
+        "changed_count": 81,
+        "total_before": 301,
+        "total_after": 264
+      }
+    }
+  ]
+}