diff --git a/internal/agent/memoryflush.go b/internal/agent/memoryflush.go
index 00c38456..8f4caad2 100644
--- a/internal/agent/memoryflush.go
+++ b/internal/agent/memoryflush.go
@@ -21,16 +21,13 @@ const (
 	DefaultMemoryFlushSystemPrompt = "Pre-compaction memory flush turn. " +
 		"The session is near auto-compaction; capture durable memories to disk. " +
 		"You may reply, but usually NO_REPLY is correct."
-
-	DefaultSoftThresholdTokens = 4000
 )
 
 // MemoryFlushSettings holds resolved flush config with defaults applied.
 type MemoryFlushSettings struct {
-	Enabled             bool
-	SoftThresholdTokens int
-	Prompt              string
-	SystemPrompt        string
+	Enabled      bool
+	Prompt       string
+	SystemPrompt string
 }
 
 // ResolveMemoryFlushSettings resolves flush settings from config, applying defaults.
@@ -39,10 +36,9 @@ func ResolveMemoryFlushSettings(compaction *config.CompactionConfig) *MemoryFlus
 	if compaction == nil || compaction.MemoryFlush == nil {
 		// Default: enabled
 		return &MemoryFlushSettings{
-			Enabled:             true,
-			SoftThresholdTokens: DefaultSoftThresholdTokens,
-			Prompt:              DefaultMemoryFlushPrompt,
-			SystemPrompt:        DefaultMemoryFlushSystemPrompt,
+			Enabled:      true,
+			Prompt:       DefaultMemoryFlushPrompt,
+			SystemPrompt: DefaultMemoryFlushSystemPrompt,
 		}
 	}
 
@@ -52,15 +48,11 @@ func ResolveMemoryFlushSettings(compaction *config.CompactionConfig) *MemoryFlus
 	}
 
 	settings := &MemoryFlushSettings{
-		Enabled:             true,
-		SoftThresholdTokens: DefaultSoftThresholdTokens,
-		Prompt:              DefaultMemoryFlushPrompt,
-		SystemPrompt:        DefaultMemoryFlushSystemPrompt,
+		Enabled:      true,
+		Prompt:       DefaultMemoryFlushPrompt,
+		SystemPrompt: DefaultMemoryFlushSystemPrompt,
 	}
 
-	if mf.SoftThresholdTokens > 0 {
-		settings.SoftThresholdTokens = mf.SoftThresholdTokens
-	}
 	if mf.Prompt != "" {
 		settings.Prompt = mf.Prompt
 	}
@@ -72,7 +64,8 @@ func ResolveMemoryFlushSettings(compaction *config.CompactionConfig) *MemoryFlus
 }
 
 // shouldRunMemoryFlush checks whether a memory flush should run before compaction.
-// Matching TS memory-flush.ts:shouldRunMemoryFlush.
+// Flush always runs when compaction triggers (called inside maybeSummarize),
+// gated only by enabled/memory checks and a dedup guard per compaction cycle.
 func (l *Loop) shouldRunMemoryFlush(sessionKey string, totalTokens int, settings *MemoryFlushSettings) bool {
 	if settings == nil || !settings.Enabled || !l.hasMemory {
 		return false
@@ -82,23 +75,7 @@ func (l *Loop) shouldRunMemoryFlush(sessionKey string, totalTokens int, settings
 		return false
 	}
 
-	// Threshold = contextWindow - reserveTokensFloor - softThresholdTokens
-	// When totalTokens >= threshold → flush
-	reserveFloor := 20000 // default
-	if l.compactionCfg != nil && l.compactionCfg.ReserveTokensFloor > 0 {
-		reserveFloor = l.compactionCfg.ReserveTokensFloor
-	}
-
-	threshold := l.contextWindow - reserveFloor - settings.SoftThresholdTokens
-	if threshold <= 0 {
-		return false
-	}
-
-	if totalTokens < threshold {
-		return false
-	}
-
-	// Deduplication: skip if already flushed in this compaction cycle
+	// Deduplication: skip if already flushed in this compaction cycle.
 	compactionCount := l.sessions.GetCompactionCount(sessionKey)
 	lastFlushAt := l.sessions.GetMemoryFlushCompactionCount(sessionKey)
 	if lastFlushAt >= 0 && lastFlushAt == compactionCount {
diff --git a/ui/web/src/i18n/locales/en/agents.json b/ui/web/src/i18n/locales/en/agents.json
index 4b45fc04..43029297 100644
--- a/ui/web/src/i18n/locales/en/agents.json
+++ b/ui/web/src/i18n/locales/en/agents.json
@@ -271,12 +271,14 @@
     "compaction": {
       "title": "Compaction",
       "description": "Context window compaction and memory flush settings",
-      "reserveTokensFloor": "Reserve Tokens Floor",
       "maxHistoryShare": "Max History Share (0-1)",
+      "maxHistoryShareTip": "Maximum fraction of context window for conversation history (e.g. 0.75 = 75%). Compaction triggers when exceeded.",
       "minMessages": "Min Messages",
+      "minMessagesTip": "Minimum messages before compaction can trigger, regardless of token count.",
       "keepLastMessages": "Keep Last Messages",
+      "keepLastMessagesTip": "Recent messages kept after compaction. Older messages are replaced by a summary.",
       "memoryFlush": "Memory Flush",
-      "softThreshold": "Soft Threshold (tokens)"
+      "memoryFlushTip": "Before compaction, the agent gets a turn to save important context to memory files. Also triggers Knowledge Graph extraction."
     },
     "contextPruning": {
       "title": "Context Pruning",
diff --git a/ui/web/src/i18n/locales/vi/agents.json b/ui/web/src/i18n/locales/vi/agents.json
index ca59d989..1048ebdb 100644
--- a/ui/web/src/i18n/locales/vi/agents.json
+++ b/ui/web/src/i18n/locales/vi/agents.json
@@ -270,13 +270,15 @@
     },
     "compaction": {
       "title": "Nén ngữ cảnh",
-      "description": "Cài đặt nén cửa sổ ngữ cảnh và xả bộ nhớ",
-      "reserveTokensFloor": "Sàn token dự trữ",
+      "description": "Cài đặt nén cửa sổ ngữ cảnh và ghi nhớ trước nén",
       "maxHistoryShare": "Phần lịch sử tối đa (0-1)",
+      "maxHistoryShareTip": "Tỷ lệ tối đa của cửa sổ ngữ cảnh dành cho lịch sử hội thoại (vd: 0.75 = 75%). Nén kích hoạt khi vượt quá.",
       "minMessages": "Tin nhắn tối thiểu",
+      "minMessagesTip": "Số tin nhắn tối thiểu trước khi nén có thể kích hoạt, bất kể số token.",
       "keepLastMessages": "Giữ tin nhắn cuối",
-      "memoryFlush": "Xả bộ nhớ",
-      "softThreshold": "Ngưỡng mềm (token)"
+      "keepLastMessagesTip": "Số tin nhắn gần nhất giữ lại sau khi nén. Các tin nhắn cũ hơn được thay thế bằng bản tóm tắt.",
+      "memoryFlush": "Ghi nhớ trước nén",
+      "memoryFlushTip": "Trước khi nén, agent được một lượt để lưu ngữ cảnh quan trọng vào file bộ nhớ. Cũng kích hoạt trích xuất Knowledge Graph."
     },
     "contextPruning": {
       "title": "Cắt bớt ngữ cảnh",
diff --git a/ui/web/src/i18n/locales/zh/agents.json b/ui/web/src/i18n/locales/zh/agents.json
index 61b232c7..22a520f7 100644
--- a/ui/web/src/i18n/locales/zh/agents.json
+++ b/ui/web/src/i18n/locales/zh/agents.json
@@ -270,13 +270,15 @@
     },
     "compaction": {
       "title": "压缩",
-      "description": "上下文窗口压缩和内存刷新设置",
-      "reserveTokensFloor": "保留令牌下限",
+      "description": "上下文窗口压缩和压缩前记忆设置",
       "maxHistoryShare": "最大历史份额 (0-1)",
+      "maxHistoryShareTip": "上下文窗口用于对话历史的最大比例（如0.75 = 75%）。超过时触发压缩。",
       "minMessages": "最小消息数",
+      "minMessagesTip": "压缩触发前所需的最少消息数，与令牌数无关。",
       "keepLastMessages": "保留最后消息数",
-      "memoryFlush": "内存刷新",
-      "softThreshold": "软阈值（令牌）"
+      "keepLastMessagesTip": "压缩后保留的最近消息数。旧消息被摘要替换。",
+      "memoryFlush": "压缩前记忆",
+      "memoryFlushTip": "压缩前，Agent可以将重要上下文保存到记忆文件。同时触发知识图谱提取。"
     },
     "contextPruning": {
       "title": "上下文裁剪",
diff --git a/ui/web/src/pages/agents/agent-detail/config-sections/compaction-section.tsx b/ui/web/src/pages/agents/agent-detail/config-sections/compaction-section.tsx
index 51e0f106..300882b2 100644
--- a/ui/web/src/pages/agents/agent-detail/config-sections/compaction-section.tsx
+++ b/ui/web/src/pages/agents/agent-detail/config-sections/compaction-section.tsx
@@ -23,16 +23,7 @@ export function CompactionSection({ enabled, value, onToggle, onChange }: Compac
     >
       <div className="grid grid-cols-1 gap-4 sm:grid-cols-2">
         <div className="space-y-2">
-          <InfoLabel tip="Minimum tokens reserved for the LLM response. Higher values give more room for output but reduce available context for history.">{t(`${s}.reserveTokensFloor`)}</InfoLabel>
-          <Input
-            type="number"
-            placeholder="20000"
-            value={value.reserveTokensFloor ?? ""}
-            onChange={(e) => onChange({ ...value, reserveTokensFloor: numOrUndef(e.target.value) })}
-          />
-        </div>
-        <div className="space-y-2">
-          <InfoLabel tip="Maximum fraction of the context window used for conversation history before compaction triggers (e.g. 0.75 = 75%).">{t(`${s}.maxHistoryShare`)}</InfoLabel>
+          <InfoLabel tip={t(`${s}.maxHistoryShareTip`)}>{t(`${s}.maxHistoryShare`)}</InfoLabel>
           <Input
             type="number"
             step="0.05"
@@ -41,10 +32,8 @@ export function CompactionSection({ enabled, value, onToggle, onChange }: Compac
             onChange={(e) => onChange({ ...value, maxHistoryShare: numOrUndef(e.target.value) })}
           />
         </div>
-      </div>
-      <div className="grid grid-cols-1 gap-4 sm:grid-cols-2">
         <div className="space-y-2">
-          <InfoLabel tip="Minimum number of messages in history before compaction can trigger, regardless of token usage.">{t(`${s}.minMessages`)}</InfoLabel>
+          <InfoLabel tip={t(`${s}.minMessagesTip`)}>{t(`${s}.minMessages`)}</InfoLabel>
           <Input
             type="number"
             placeholder="50"
@@ -52,8 +41,10 @@ export function CompactionSection({ enabled, value, onToggle, onChange }: Compac
             onChange={(e) => onChange({ ...value, minMessages: numOrUndef(e.target.value) })}
           />
         </div>
+      </div>
+      <div className="grid grid-cols-1 gap-4 sm:grid-cols-2">
         <div className="space-y-2">
-          <InfoLabel tip="Number of recent messages to preserve after compaction. The rest is replaced by a summary.">{t(`${s}.keepLastMessages`)}</InfoLabel>
+          <InfoLabel tip={t(`${s}.keepLastMessagesTip`)}>{t(`${s}.keepLastMessages`)}</InfoLabel>
           <Input
             type="number"
             placeholder="4"
@@ -62,36 +53,14 @@ export function CompactionSection({ enabled, value, onToggle, onChange }: Compac
           />
         </div>
       </div>
-      <div className="space-y-3">
-        <div className="flex items-center gap-2">
-          <Switch
-            checked={value.memoryFlush?.enabled ?? true}
-            onCheckedChange={(v) =>
-              onChange({ ...value, memoryFlush: { ...value.memoryFlush, enabled: v } })
-            }
-          />
-          <InfoLabel tip="When enabled, compacted history is also saved to long-term memory for future retrieval.">{t(`${s}.memoryFlush`)}</InfoLabel>
-        </div>
-        {(value.memoryFlush?.enabled ?? true) && (
-          <div className="space-y-2 pl-6">
-            <InfoLabel tip="Token count threshold that triggers memory flush. When summary exceeds this, older memories are flushed to storage.">{t(`${s}.softThreshold`)}</InfoLabel>
-            <Input
-              type="number"
-              placeholder="4000"
-              value={value.memoryFlush?.softThresholdTokens ?? ""}
-              onChange={(e) =>
-                onChange({
-                  ...value,
-                  memoryFlush: {
-                    ...value.memoryFlush,
-                    enabled: true,
-                    softThresholdTokens: numOrUndef(e.target.value),
-                  },
-                })
-              }
-            />
-          </div>
-        )}
+      <div className="flex items-center gap-2">
+        <Switch
+          checked={value.memoryFlush?.enabled ?? true}
+          onCheckedChange={(v) =>
+            onChange({ ...value, memoryFlush: { ...value.memoryFlush, enabled: v } })
+          }
+        />
+        <InfoLabel tip={t(`${s}.memoryFlushTip`)}>{t(`${s}.memoryFlush`)}</InfoLabel>
       </div>
     </ConfigSection>
   );