post-compaction 추적과 대화별 사용량 계측 보강

- claude-code의 post-compaction 플래그 흐름을 참고해 compact 직후 첫 응답을 별도로 추적하고 대화 단위 후속 사용량을 집계함 - ChatConversation에 pending post-compaction, 응답 회수, prompt/completion token 누적 필드를 추가하고 AX Agent 컨텍스트 카드 hover에 표시함 - README 및 docs/DEVELOPMENT.md에 2026-04-04 23:47 (KST) 기준 이력을 반영함 - 검증: dotnet build src/AxCopilot/AxCopilot.csproj -c Release -v minimal -p:OutputPath=bin\\verify\\ -p:IntermediateOutputPath=obj\\verify\\ (경고 0 / 오류 0)
2026-04-04 23:54:59 +09:00
parent d2f8e39d2b
commit ca006972b2
4 changed files with 59 additions and 2 deletions
--- a/README.md
+++ b/README.md
@@ -12,6 +12,8 @@ Windows 전용 시맨틱 런처 & 워크스페이스 매니저
 - AX Agent 컨텍스트 압축 경로에 `session memory compact`, `microcompact`, `collapse/snip` 단계를 추가해 오래된 요약·실행 로그·도구 결과를 LLM 요약 전에 더 세밀하게 줄이도록 보강했습니다.
 - 현재 대화 기준 compact 누적 회수, 자동/수동 비중, 절감 토큰, session memory 적용 횟수, microcompact/snipped 메시지 수를 하단 컨텍스트 카드 hover에서 함께 확인할 수 있게 했습니다.
 - 검증: `dotnet build src/AxCopilot/AxCopilot.csproj -c Release -v minimal -p:OutputPath=bin\\verify\\ -p:IntermediateOutputPath=obj\\verify\\` 경고 0 / 오류 0
+- compact 직후 첫 응답을 별도로 추적하는 post-compaction 흐름을 추가해, 압축 다음 턴의 응답 횟수와 사용 토큰을 현재 대화 기준으로 누적 집계하도록 보강했습니다.
+- AX Agent 하단 컨텍스트 카드 hover에는 이제 `compact 후 첫 응답 대기 중`, `compact 이후 응답 수`, `compact 이후 사용 토큰`까지 함께 표시됩니다.


 - 업데이트: 2026-04-04 23:32 (KST)
--- a/docs/DEVELOPMENT.md
+++ b/docs/DEVELOPMENT.md
@@ -2,6 +2,7 @@

 - Document update: 2026-04-04 23:47 (KST) - Added AX-side session-memory compaction, collapse/snip trimming, and per-conversation compaction usage tracking so compact now records staged reductions instead of only the last before/after snapshot.
 - Document update: 2026-04-04 23:47 (KST) - The AX Agent context tooltip now shows session totals for compaction count, saved tokens, session-memory passes, microcompact boundaries, and snipped messages for the current conversation.
+- Document update: 2026-04-04 23:47 (KST) - Added a post-compaction flag and conversation-level follow-up usage counters so AX can distinguish the first response after compaction and report its prompt/completion token cost separately.

 - Document update: 2026-04-04 23:32 (KST) - Upgraded ContextCondenser from a two-step path to a three-step path (tool-result truncate -> microcompact -> summarize) so long sessions can strip older execution/tool chatter before invoking the LLM summary stage.
 - Document update: 2026-04-04 23:32 (KST) - Added AX-side microcompact boundaries that compress older tool results, execution metadata, and oversized messages into lighter boundary summaries, bringing the compact flow closer to the staged approach used in claude-code.
--- a/src/AxCopilot/Models/ChatModels.cs
+++ b/src/AxCopilot/Models/ChatModels.cs
@@ -175,6 +175,18 @@ public class ChatConversation

    [JsonPropertyName("lastCompactionStageSummary")]
    public string? LastCompactionStageSummary { get; set; }
+
+    [JsonPropertyName("pendingPostCompaction")]
+    public bool PendingPostCompaction { get; set; }
+
+    [JsonPropertyName("postCompactionResponseCount")]
+    public int PostCompactionResponseCount { get; set; }
+
+    [JsonPropertyName("postCompactionPromptTokens")]
+    public int PostCompactionPromptTokens { get; set; }
+
+    [JsonPropertyName("postCompactionCompletionTokens")]
+    public int PostCompactionCompletionTokens { get; set; }
 }

 public class ChatAgentRunRecord
--- a/src/AxCopilot/Views/ChatWindow.xaml.cs
+++ b/src/AxCopilot/Views/ChatWindow.xaml.cs
@@ -107,6 +107,10 @@ public partial class ChatWindow : Window
    private int _sessionMemoryCompactionCount;
    private int _sessionMicrocompactBoundaryCount;
    private int _sessionSnipCompactionCount;
+    private bool _pendingPostCompaction;
+    private int _sessionPostCompactionResponseCount;
+    private int _sessionPostCompactionPromptTokens;
+    private int _sessionPostCompactionCompletionTokens;
    private void ApplyQuickActionVisual(Button button, bool active, string activeBg, string activeFg)
    {
        if (button?.Content is not string text)
@@ -1732,6 +1736,10 @@ public partial class ChatWindow : Window
        _sessionMemoryCompactionCount = conv?.SessionMemoryCompactionCount ?? 0;
        _sessionMicrocompactBoundaryCount = conv?.MicrocompactBoundaryCount ?? 0;
        _sessionSnipCompactionCount = conv?.SnipCompactionCount ?? 0;
+        _pendingPostCompaction = conv?.PendingPostCompaction ?? false;
+        _sessionPostCompactionResponseCount = conv?.PostCompactionResponseCount ?? 0;
+        _sessionPostCompactionPromptTokens = conv?.PostCompactionPromptTokens ?? 0;
+        _sessionPostCompactionCompletionTokens = conv?.PostCompactionCompletionTokens ?? 0;
        _lastCompactionAt = conv?.LastCompactionAt;
        _lastCompactionWasAutomatic = conv?.LastCompactionWasAutomatic ?? false;
        _lastCompactionBeforeTokens = conv?.LastCompactionBeforeTokens;
@@ -10939,10 +10947,12 @@ public partial class ChatWindow : Window
            ? _agentCumulativeOutputTokens
            : usage?.CompletionTokens ?? 0;

+        var wasPostCompactionResponse = _pendingPostCompaction;
        if (displayInput > 0 || displayOutput > 0)
        {
            UpdateStatusTokens(displayInput, displayOutput);
            Services.UsageStatisticsService.RecordTokens(displayInput, displayOutput);
+            ConsumePostCompactionUsageIfNeeded(displayInput, displayOutput);
        }
        string tokenText;
        if (displayInput > 0 || displayOutput > 0)
@@ -10952,9 +10962,10 @@ public partial class ChatWindow : Window
        else
            tokenText = $"~{FormatTokenCount(EstimateTokenCount(finalContent))} tokens";

+        var postCompactSuffix = wasPostCompactionResponse ? "  ·  compact 직후" : "";
        var metaText = new TextBlock
        {
-            Text = $"{elapsedText}  ·  {tokenText}",
+            Text = $"{elapsedText}  ·  {tokenText}{postCompactSuffix}",
            FontSize = 9.5,
            Foreground = TryFindResource("SecondaryText") as Brush ?? Brushes.Gray,
            HorizontalAlignment = HorizontalAlignment.Right,
@@ -16532,6 +16543,11 @@ private static (string icon, string label, string bgHex, string fgHex) GetDecisi
              $"세션 절감: {Services.TokenEstimator.Format(_sessionCompactionSavedTokens)} tokens\n" +
              $"세션 메모리 {_sessionMemoryCompactionCount:N0}회 · 경계 {_sessionMicrocompactBoundaryCount:N0}건 · snip {_sessionSnipCompactionCount:N0}건"
            : "";
+        var postCompactionUsage = _sessionPostCompactionResponseCount > 0
+            ? $"\ncompact 이후 응답: {_sessionPostCompactionResponseCount:N0}회\n" +
+              $"compact 이후 사용량: {Services.TokenEstimator.Format(_sessionPostCompactionPromptTokens)} + {Services.TokenEstimator.Format(_sessionPostCompactionCompletionTokens)} = {Services.TokenEstimator.Format(_sessionPostCompactionPromptTokens + _sessionPostCompactionCompletionTokens)} tokens"
+            : "";
+        var pendingPostCompaction = _pendingPostCompaction ? "\ncompact 후 첫 응답 대기 중" : "";

        TokenUsageCard.ToolTip =
            $"상태: {summary}\n" +
@@ -16540,7 +16556,9 @@ private static (string icon, string label, string bgHex, string fgHex) GetDecisi
            $"자동 압축 시작: {triggerPercent}%\n" +
            $"현재 입력 초안 포함" +
            compactHistory +
-            compactSession;
+            compactSession +
+            postCompactionUsage +
+            pendingPostCompaction;

        UpdateCircularUsageArc(TokenUsageArc, usageRatio, 18, 18, 14);
        PositionThresholdMarker(TokenUsageThresholdMarker, triggerRatio, 18, 18, 14, 3);
@@ -16605,6 +16623,7 @@ private static (string icon, string label, string bgHex, string fgHex) GetDecisi
        _sessionMemoryCompactionCount += result.SessionMemoryApplied ? 1 : 0;
        _sessionMicrocompactBoundaryCount += result.MicrocompactBoundaryCount;
        _sessionSnipCompactionCount += result.SnippedMessageCount + result.CollapsedBoundaryCount;
+        _pendingPostCompaction = true;
        if (wasAutomatic)
            _sessionAutomaticCompactionCount++;
        else
@@ -16627,6 +16646,29 @@ private static (string icon, string label, string bgHex, string fgHex) GetDecisi
        conv.LastCompactionBeforeTokens = beforeTokens;
        conv.LastCompactionAfterTokens = afterTokens;
        conv.LastCompactionStageSummary = _lastCompactionStageSummary;
+        conv.PendingPostCompaction = true;
+        try { _storage.Save(conv); } catch { }
+    }
+
+    private void ConsumePostCompactionUsageIfNeeded(int promptTokens, int completionTokens)
+    {
+        if (!_pendingPostCompaction)
+            return;
+
+        _pendingPostCompaction = false;
+        _sessionPostCompactionResponseCount++;
+        _sessionPostCompactionPromptTokens += Math.Max(0, promptTokens);
+        _sessionPostCompactionCompletionTokens += Math.Max(0, completionTokens);
+
+        ChatConversation? conv;
+        lock (_convLock) conv = _currentConversation;
+        if (conv == null)
+            return;
+
+        conv.PendingPostCompaction = false;
+        conv.PostCompactionResponseCount = _sessionPostCompactionResponseCount;
+        conv.PostCompactionPromptTokens = _sessionPostCompactionPromptTokens;
+        conv.PostCompactionCompletionTokens = _sessionPostCompactionCompletionTokens;
        try { _storage.Save(conv); } catch { }
    }