From e83873e822a93d44b228add3c7a9de333f6ca8d7 Mon Sep 17 00:00:00 2001
From: CC Worker <cc-worker@local>
Date: Mon, 8 Jun 2026 18:02:51 +0000
Subject: [PATCH] fix(exam): dedupe all AI auto-map rows by id before insert

B1-4 live-route validation: continuation bands re-emit the same stable AI id for
response_areas/boundaries/layout (not just questions), causing duplicate-pkey insert
failures. Add _dedupe_rows_by_id applied to all four tables in _refresh_ai_rows.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---
 routers/exam/templates.py    | 17 ++++++++++++++++-
 tests/test_exam_templates.py | 14 ++++++++++++++
 2 files changed, 30 insertions(+), 1 deletion(-)

diff --git a/routers/exam/templates.py b/routers/exam/templates.py
index 736b35f..1090751 100644
--- a/routers/exam/templates.py
+++ b/routers/exam/templates.py
@@ -543,12 +543,27 @@ def _map_first_pass_to_rows(template_id: str, first_pass: Dict[str, Any], pdf_by
     return {"questions": questions, "response_areas": response_areas, "boundaries": boundaries, "layout": layout}
 
 
+def _dedupe_rows_by_id(rows: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+    """Preserve first occurrence of stable AI row ids emitted by noisy OCR detectors."""
+    out: List[Dict[str, Any]] = []
+    seen: set[str] = set()
+    for row in rows:
+        row_id = row.get("id")
+        if row_id:
+            key = str(row_id)
+            if key in seen:
+                continue
+            seen.add(key)
+        out.append(row)
+    return out
+
+
 def _refresh_ai_rows(ctx: ExamContext, template_id: str, rows: Dict[str, List[Dict[str, Any]]]) -> None:
     sb = ctx.supabase
     for table in ("exam_response_areas", "exam_boundaries", "exam_template_layout", "exam_questions"):
         sb.table(table).delete().eq("template_id", template_id).eq("source", "ai").eq("confirmed", False).execute()
     for table, key in (("exam_questions", "questions"), ("exam_response_areas", "response_areas"), ("exam_boundaries", "boundaries"), ("exam_template_layout", "layout")):
-        payload = rows.get(key) or []
+        payload = _dedupe_rows_by_id(rows.get(key) or [])
         if payload:
             sb.table(table).insert(payload).execute()
 
diff --git a/tests/test_exam_templates.py b/tests/test_exam_templates.py
index 307e9ef..37e38d2 100644
--- a/tests/test_exam_templates.py
+++ b/tests/test_exam_templates.py
@@ -642,6 +642,20 @@ def test_auto_map_fast_path_merges_ai_rows_and_returns_detail(monkeypatch):
     assert store["exam_boundaries"] and store["exam_boundaries"][0]["derivation"] == "docling-main-band"
 
 
+def test_auto_map_deduplicates_repeated_response_area_ids(monkeypatch):
+    store = _template_with_source()
+    client, store = make_client(store=store)
+    _patch_auto_map(monkeypatch, store, fast=True)
+    dup = {"page_index": 0, "bbox": {"l": 50, "t": 700, "r": 100, "b": 680, "coord_origin": "BOTTOMLEFT"}, "region_type": "answer_lines", "confidence": 0.9}
+    monkeypatch.setattr(templates_mod, "detect_response_regions_from_pdf", lambda *_a, **_k: [dup, dict(dup)])
+
+    resp = client.post("/api/exam/templates/t1/auto-map")
+
+    assert resp.status_code == 200
+    response_area_ids = [r["id"] for r in store["exam_response_areas"]]
+    assert len(response_area_ids) == len(set(response_area_ids))
+
+
 def test_auto_map_preserves_manual_and_confirmed_rows_on_rerun(monkeypatch):
     store = _template_with_source()
     store.update({