liuwei
f593f5dd90
优化 ai_auto_response 模型侧定向辱骂响应策略
- 增加 directed abuse 场景识别，只作为模型输入信号，不做本地硬编码回复
- 在触发与规划层为定向挑衅场景单独标记，并强制走 social_short 短回复模式
- 将 abuse_directed 信号写入 Dify control，帮助模型稳定识别被点名挑衅场景
- 优化 Dify 主提示词与保守降级提示词，要求 abuse_directed 时默认短回且不要空掉
- 保持回复仍由模型生成，避免本地模板化回复暴露机器人痕迹
2026-04-24 14:44:33 +08:00
..
2026-04-24 14:44:33 +08:00
2026-01-16 13:34:37 +08:00
2026-01-16 13:34:37 +08:00
2026-04-17 09:42:00 +08:00
2026-04-17 09:42:00 +08:00
2026-04-20 14:45:03 +08:00
2026-04-23 15:53:46 +08:00
2026-04-22 09:43:47 +08:00
2026-01-16 13:34:37 +08:00
2026-04-17 09:42:00 +08:00
2026-04-22 12:34:26 +08:00
2026-04-23 14:16:55 +08:00
2026-04-20 14:45:03 +08:00
2026-04-22 11:40:17 +08:00
2026-04-20 14:45:03 +08:00
2026-01-16 13:34:37 +08:00
2026-04-20 10:42:46 +08:00
2026-01-16 13:34:37 +08:00
2026-01-22 11:59:21 +08:00
2026-01-26 15:28:54 +08:00
2026-01-20 17:08:31 +08:00
2026-01-16 13:34:37 +08:00
2026-01-16 13:34:37 +08:00
2026-04-20 14:36:56 +08:00
2026-02-26 09:17:56 +08:00
2026-01-16 13:35:45 +08:00
2026-04-24 09:02:09 +08:00
2026-01-26 15:28:54 +08:00
2025-03-27 16:12:38 +08:00
2026-01-16 13:34:37 +08:00
2026-04-20 11:44:26 +08:00
2026-04-20 13:12:44 +08:00
2026-04-21 15:32:21 +08:00
2026-01-16 13:34:37 +08:00
2026-01-19 14:15:46 +08:00
2026-04-21 16:39:21 +08:00
2026-04-21 15:52:51 +08:00
2026-02-25 15:36:21 +08:00
2026-01-16 13:34:37 +08:00
2026-04-16 15:49:02 +08:00
2026-04-16 16:05:59 +08:00
2026-01-16 13:34:37 +08:00