tune xiaoniu reply brevity and flow thresholds

plugins/ai_auto_response/config.toml

@@ -15,8 +15,8 @@ endpoint = "chat/completions"
 api_key = "sk-XTWwXIgo2QMyX8AwBg0NQrxaDkvQiCX8rfylfmnHID5zdjMt"
 model = "gpt-5.4"
 timeout_seconds = 45
-temperature = 0.7
-max_tokens = 500
+temperature = 0.35
+max_tokens = 120
 stream = true
 
 [mode]
@@ -39,24 +39,24 @@ casual_topic = 0.35
 
 [flow]
 enable_flow_state = true
-flow_decay_per_minute = 8
-idle_threshold = 20
-warming_threshold = 40
-engaged_threshold = 70
+flow_decay_per_minute = 12
+idle_threshold = 30
+warming_threshold = 55
+engaged_threshold = 90
 at_bot_boost = 40
 question_boost = 30
 followup_boost = 20
-topic_boost = 15
-returning_member_boost = 10
-response_accepted_boost = 15
+topic_boost = 8
+returning_member_boost = 6
+response_accepted_boost = 10
 ignored_reply_penalty = 20
-over_reply_penalty = 15
+over_reply_penalty = 22
 night_penalty = 30
-max_bot_reply_streak = 3
+max_bot_reply_streak = 2
 
 [cooldown]
-group_reply_cooldown_sec = 45
-same_user_followup_cooldown_sec = 10
+group_reply_cooldown_sec = 90
+same_user_followup_cooldown_sec = 18
 night_silent_hours = ["01:00-07:30"]
 
 [memory]
@@ -69,6 +69,7 @@ qdrant_collection = "abot_xiaoniu_memory"
 ollama_base_url = "http://192.168.2.50:11434"
 embedding_model = "bge-m3"
 vector_top_k = 5
+max_context_memories = 2
 vector_min_score = 0.65
 vector_trigger_modes = ["returning_member", "long_absent_member", "qa_with_context", "reactivated_topic"]