同步代码更新 (2026-06-16)

- 移除 generate_diverse_answers.py（多样化回答已废止）
- 更新 generate_fresh_qa.py（目标50万条新问答）
- 更新 generate_multiturn.py（6~12轮长对话，目标50万）
- 更新 regenerate_all_high_temp.py（独立保存，不追加训练集）
- 更新 .gitignore（忽略新输出目录）
- 所有API Key改为环境变量读取

regenerate_all_high_temp.py

@@ -111,30 +111,12 @@ def main():
     print(f"生成结果: {len(all_results)} 条 (去重后)")
     print(f"用时: {(time.time() - t_start)/60:.1f} 分钟")
     
-    # 5. 追加到训练集
-    for tgt in ['xiaoli_train_v2.json', 'xiaoli_merged_data_v2.json']:
-        tp = os.path.join(MODEL_DIR, tgt)
-        print(f"\n追加到 {tgt}...")
-        with open(tp, encoding='utf-8') as f:
-            cur = json.load(f)
-        
-        # 去重
-        ex = set((it.get('question', '')[:100], it.get('answer', '')[:100]) for it in cur)
-        to_add = []
-        for r in all_results:
-            key = (r['question'][:100], r['answer'][:100])
-            if key not in ex:
-                ex.add(key)
-                to_add.append({'question': r['question'], 'answer': r['answer']})
-        
-        cur.extend(to_add)
-        with open(tp, 'w', encoding='utf-8') as f:
-            json.dump(cur, f, ensure_ascii=False, indent=2)
-        
-        if 'train' in tgt:
-            print(f"  ✅ 训练集 +{len(to_add)} 条 → {len(cur)} 条")
-        else:
-            print(f"  ✅ 合并集 +{len(to_add)} 条 → {len(cur)} 条")
+    # 5. 保存最终去重结果（独立存放，不追加到训练集）
+    final_file = os.path.join('/data/user/work', 'high_temp_final.json')
+    with open(final_file, 'w', encoding='utf-8') as f:
+        json.dump(all_results, f, ensure_ascii=False, indent=2)
+    print(f"\n最终结果已保存: {final_file}")
+    print(f"如需使用，可手动合并到训练集")
     
     print(f"\n{'='*60}")
     print(f"全部完成! 新增 {len(all_results)} 条多样化回答")