From d0194f5840518aac9c38f078fff2fe2129471db9 Mon Sep 17 00:00:00 2001
From: iomgaa <iomgaaycz@gmail.com>
Date: Tue, 14 Jul 2026 14:11:13 -0400
Subject: [PATCH] fix: degrade distractor pool gracefully on malformed VLM
 response

---
 app/question_gen/distractor_selector.py | 15 ++++++++++-----
 tests/unit/test_distractor_selector.py  |  8 ++++++++
 2 files changed, 18 insertions(+), 5 deletions(-)

diff --git a/app/question_gen/distractor_selector.py b/app/question_gen/distractor_selector.py
index 6e3bf2a..dc2ce9e 100644
--- a/app/question_gen/distractor_selector.py
+++ b/app/question_gen/distractor_selector.py
@@ -53,7 +53,7 @@ class SelectorOutcome:
         answer: 正解字母（恒 "A"），hard-fail 时为 None。
     """
 
-    observation: dict
+    observation: dict[str, object]
     options: tuple[str, ...] | None = None
     answer: str | None = None
 
@@ -142,7 +142,12 @@ async def _generate_pool(
     user = _material_context_block(question, correct_text, material) + f"\n## N\nGenerate exactly {n} distractors."
     messages = [{"role": "system", "content": system}, {"role": "user", "content": user}]
     resp = await vlm.chat_with_images(messages, list(material.frame_paths), session_id=session_id)
-    data = _parse_json_object(resp.content)
+    # graceful 降级：解析爆炸时返回空池，让退火/hard_fail 接管（区别于契约违反的打分校验）
+    try:
+        data = _parse_json_object(resp.content)
+    except (ValueError, json.JSONDecodeError) as exc:
+        logger.warning("grounded selector 候选池响应解析失败，降级为空池: {}", exc)
+        return []
     raw = data.get("distractors", [])
     if not isinstance(raw, list):
         return []
@@ -187,7 +192,7 @@ async def build_grounded_options(
 ) -> SelectorOutcome:
     """生成候选池 → 视觉打分 → 区间选 3 干扰项 → 重组四选项。
 
-    退火（凑不齐 3 个时按序）：① 追加 N 个候选使池达 2N 再打分；② 逐步放宽
+    退火（凑不齐 3 个时按序）：① 最多追加 N 个新候选（去重后实际增量可能更少）再打分；② 逐步放宽
     δ_high（纯重选，不再调 VLM）；③ 仍不足则 hard_fail（调用方走重出）。
 
     参数:
@@ -214,7 +219,7 @@ async def build_grounded_options(
         correct_score, candidates, cand_scores, config.delta_low, config.delta_high
     )
 
-    # 退火 1: 追加 N 个候选使池达 2N（仅对新增候选打分，正解分保持首轮值）
+    # 退火 1: 最多追加 N 个新候选（去重后实际增量可能更少），仅对新增候选打分，正解分保持首轮值
     if chosen is None:
         anneal_rounds += 1
         more = await _generate_pool(
@@ -243,7 +248,7 @@ async def build_grounded_options(
         )
 
     hard_fail = chosen is None
-    observation = {
+    observation: dict[str, object] = {
         "correct_score": correct_score,
         "chosen": [
             cand_scores[candidates.index(c)] for c in (chosen or [])
diff --git a/tests/unit/test_distractor_selector.py b/tests/unit/test_distractor_selector.py
index cf340f0..95edc3f 100644
--- a/tests/unit/test_distractor_selector.py
+++ b/tests/unit/test_distractor_selector.py
@@ -55,6 +55,14 @@ class _Material:
     source_nodes = ("n1",)
 
 
+@pytest.mark.asyncio
+async def test_generate_pool_degrades_on_malformed_response():
+    from app.question_gen.distractor_selector import _generate_pool
+    vlm = _FakeVLM(['["not", "a", "dict"]'])  # 顶层是 list 不是 dict
+    out = await _generate_pool(vlm, "?", "蒸", _Material(), 4, session_id="s")
+    assert out == []
+
+
 @pytest.mark.asyncio
 async def test_build_grounded_options_happy_path():
     pool = '{"distractors": ["炒", "煮", "炸", "烤"]}'