fix(core): 脚本 agent 兼容模型 ScriptDraft/shots 变体,修旁白/画面全空

豆包 Doubao-Seed-2.0-Lite 常不按扁平契约输出,而是回
{"ScriptDraft":{"basicInfo":...,"shots":[{scene,dialogue,subtitle}]}}。
normalize_draft 在顶层找不到 segments → 走补占位镜兜底 → 旁白/画面全空(只剩 role)。

normalize_draft 增加容错:
- 解开 {"ScriptDraft":{...}} 外壳;basicInfo 的 camelCase 时长/比例拍平
- shots → segments;每镜 scene→画面、dialogue/subtitle→旁白
- 防 dialogue 为整句字符串时被逐字符遍历
原扁平契约路径不受影响。新增 apps/ai/tests.py 3 个回归用例,全套 21 测试通过。

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

core/backend/apps/ai/script_agent.py

@@ -220,6 +220,25 @@ def normalize_draft(raw_text: str, *, aspect_ratio: str, total_duration: int) ->
     if not isinstance(draft, dict):
         raise ValueError("脚本 JSON 顶层不是对象")
 
+    # 兼容模型不按契约的常见变体:① 把内容裹进 {"ScriptDraft": {...}} 外壳;
+    # ② 用 basicInfo(camelCase)放时长/比例;③ 用 shots 代替 segments。
+    # 解开外壳并把别名拍平到契约字段,避免「找不到 segments → 全填空占位镜」(旁白/画面全空)。
+    for wrapper in ("ScriptDraft", "script_draft", "scriptDraft", "draft"):
+        inner = draft.get(wrapper)
+        if isinstance(inner, dict):
+            draft = {**inner, **{k: v for k, v in draft.items() if k != wrapper}}
+            break
+    basic = draft.get("basicInfo") if isinstance(draft.get("basicInfo"), dict) else {}
+    if basic:
+        draft.setdefault("total_duration", basic.get("totalDuration") or basic.get("total_duration"))
+        draft.setdefault("aspect_ratio", basic.get("aspectRatio") or basic.get("aspect_ratio"))
+        draft.setdefault("theme", basic.get("theme"))
+    if not isinstance(draft.get("segments"), list):
+        for alias in ("shots", "scenes_list", "shotList"):
+            if isinstance(draft.get(alias), list):
+                draft["segments"] = draft[alias]
+                break
+
     draft["aspect_ratio"] = (draft.get("aspect_ratio") or aspect_ratio or "9:16").strip()
     dur = _nearest_duration(draft.get("total_duration") or total_duration)
     draft["total_duration"] = dur
@@ -268,19 +287,29 @@ def normalize_draft(raw_text: str, *, aspect_ratio: str, total_duration: int) ->
         speaker = seg.get("speaker")
         speaker = speaker if (speaker in valid_ids) else None
         refs = [r for r in (seg.get("entity_refs") or []) if r in valid_ids]
-        # 对白(剧情向):[{speaker(合法 entity id 或 null=旁白), line}];默认空 = 纯口播
+        # 对白(剧情向):[{speaker(合法 entity id 或 null=旁白), line}];默认空 = 纯口播。
+        # 模型变体里 dialogue 可能是字符串(整句口播)而非数组——只在数组时按结构解析。
+        raw_dialogue = seg.get("dialogue")
         dialogue = []
-        for d in (seg.get("dialogue") or []):
-            if not isinstance(d, dict):
-                continue
-            line = (d.get("line") or d.get("text") or "").strip()
-            if not line:
-                continue
-            sp = d.get("speaker")
-            dialogue.append({"speaker": sp if sp in valid_ids else None, "line": line})
+        if isinstance(raw_dialogue, list):
+            for d in raw_dialogue:
+                if not isinstance(d, dict):
+                    continue
+                line = (d.get("line") or d.get("text") or "").strip()
+                if not line:
+                    continue
+                sp = d.get("speaker")
+                dialogue.append({"speaker": sp if sp in valid_ids else None, "line": line})
+        # 旁白:narration > 字符串 dialogue(整句口播)> subtitle(字幕)> 结构化对白拼接
         narration = (seg.get("narration") or "").strip()
+        if not narration and isinstance(raw_dialogue, str):
+            narration = raw_dialogue.strip()
+        if not narration:
+            narration = (seg.get("subtitle") or "").strip()
         if not narration and dialogue:
             narration = " ".join(d["line"] for d in dialogue)  # 扁平拼接,兼容下游字幕/配音
+        # 画面:visual > visual_prompt > scene(模型变体常用 scene 描述画面)
+        visual = (seg.get("visual") or seg.get("visual_prompt") or seg.get("scene") or "").strip()
         norm_segments.append(
             {
                 "index": i,
@@ -288,7 +317,7 @@ def normalize_draft(raw_text: str, *, aspect_ratio: str, total_duration: int) ->
                 "role": role,
                 "narration": narration,
                 "speaker": speaker,
-                "visual": (seg.get("visual") or seg.get("visual_prompt") or "").strip(),
+                "visual": visual,
                 "product_exposure": (seg.get("product_exposure") or "").strip(),
                 "entity_refs": refs,
                 "dialogue": dialogue,