perf: reuse preflight transcript scan size

steipete · steipete · commit 8d5f6c8ae465 · 2026-05-28T08:31:06.000+01:00
diff --git a/src/auto-reply/reply/agent-runner-memory.test.ts b/src/auto-reply/reply/agent-runner-memory.test.ts
@@ -1614,7 +1614,7 @@ describe("runMemoryFlushIfNeeded", () => {
             usage: { input: 40_000, output: 2_000 },
           },
         }),
-      ].join("\n"),
+      ].join("\n") + "\n",
       "utf8",
     );
     registerMemoryFlushPlanResolverForTest(() => ({
@@ -1631,36 +1631,50 @@ describe("runMemoryFlushIfNeeded", () => {
       updatedAt: Date.now(),
       totalTokensFresh: false,
     };
+    const originalStat = fsCore.promises.stat.bind(fsCore.promises);
+    const statSpy = vi
+      .spyOn(fsCore.promises, "stat")
+      .mockImplementation(async (target, options) => originalStat(target, options));
 
-    const entry = await runPreflightCompactionIfNeeded({
-      cfg: {
-        agents: {
-          defaults: {
-            compaction: {
-              memoryFlush: {},
-              truncateAfterCompaction: true,
-              maxActiveTranscriptBytes: "10mb",
+    let entry: SessionEntry | undefined;
+    let directTranscriptStats: unknown[] = [];
+    try {
+      entry = await runPreflightCompactionIfNeeded({
+        cfg: {
+          agents: {
+            defaults: {
+              compaction: {
+                memoryFlush: {},
+                truncateAfterCompaction: true,
+                maxActiveTranscriptBytes: "10mb",
+              },
             },
           },
         },
-      },
-      followupRun: createTestFollowupRun({
-        sessionId: "session",
-        sessionFile,
+        followupRun: createTestFollowupRun({
+          sessionId: "session",
+          sessionFile,
+          sessionKey: "main",
+        }),
+        defaultModel: "anthropic/claude-opus-4-6",
+        agentCfgContextTokens: 100_000,
+        sessionEntry,
+        sessionStore: { main: sessionEntry },
         sessionKey: "main",
-      }),
-      defaultModel: "anthropic/claude-opus-4-6",
-      agentCfgContextTokens: 100_000,
-      sessionEntry,
-      sessionStore: { main: sessionEntry },
-      sessionKey: "main",
-      storePath: path.join(rootDir, "sessions.json"),
-      isHeartbeat: false,
-      replyOperation: createReplyOperation(),
-    });
+        storePath: path.join(rootDir, "sessions.json"),
+        isHeartbeat: false,
+        replyOperation: createReplyOperation(),
+      });
+      directTranscriptStats = statSpy.mock.calls.filter(
+        ([target]) => String(target) === sessionFile,
+      );
+    } finally {
+      statSpy.mockRestore();
+    }
 
     expect(entry).toBe(sessionEntry);
     expect(compactEmbeddedAgentSessionMock).not.toHaveBeenCalled();
+    expect(directTranscriptStats).toEqual([]);
   });
 
   it("triggers preflight compaction when the active transcript exceeds the configured byte threshold", async () => {
diff --git a/src/auto-reply/reply/agent-runner-memory.ts b/src/auto-reply/reply/agent-runner-memory.ts
@@ -571,6 +571,7 @@ function estimatePostUsageTrailingBytes(lines: string[]): number {
 type TranscriptTokenEstimate = {
   promptTokens: number;
   outputTokens?: number;
+  transcriptByteSize?: number;
   transcriptBytesTokens?: number;
 };
 
@@ -620,6 +621,7 @@ async function estimatePromptTokensFromSessionTranscript(params: {
       return {
         promptTokens: Math.ceil(promptTokens),
         outputTokens: Math.ceil(outputTokens),
+        transcriptByteSize: snapshot.byteSize,
         transcriptBytesTokens,
       };
     }
@@ -648,6 +650,7 @@ async function estimatePromptTokensFromSessionTranscript(params: {
           typeof outputTokens === "number" && Number.isFinite(outputTokens) && outputTokens > 0
             ? Math.ceil(outputTokens)
             : undefined,
+        transcriptByteSize: snapshot.byteSize,
         transcriptBytesTokens,
       };
     }
@@ -657,6 +660,7 @@ async function estimatePromptTokensFromSessionTranscript(params: {
     }
     return {
       promptTokens: Math.ceil(estimatedTokens),
+      transcriptByteSize: snapshot.byteSize,
       transcriptBytesTokens,
     };
   } catch {
@@ -740,29 +744,11 @@ export async function runPreflightCompactionIfNeeded(params: {
     typeof persistedTotalTokens === "number" &&
     Number.isFinite(persistedTotalTokens) &&
     persistedTotalTokens > 0;
-  const maxActiveTranscriptBytes = resolveMaxActiveTranscriptBytes(params.cfg);
-  const shouldCheckActiveTranscriptBytes = typeof maxActiveTranscriptBytes === "number";
-  const transcriptSizeSnapshot = shouldCheckActiveTranscriptBytes
-    ? await readSessionLogSnapshot({
-        sessionId: entry.sessionId,
-        sessionEntry:
-          entry.sessionFile || !params.followupRun.run.sessionFile
-            ? entry
-            : { ...entry, sessionFile: params.followupRun.run.sessionFile },
-        sessionKey: params.sessionKey ?? params.followupRun.run.sessionKey,
-        opts: { storePath: params.storePath },
-        includeByteSize: true,
-        includeUsage: false,
-      })
-    : undefined;
-  const activeTranscriptBytes = transcriptSizeSnapshot?.byteSize;
-  const shouldCompactByTranscriptBytes =
-    typeof activeTranscriptBytes === "number" &&
-    typeof maxActiveTranscriptBytes === "number" &&
-    activeTranscriptBytes >= maxActiveTranscriptBytes;
   const promptTokenEstimate = estimatePromptTokensForMemoryFlush(
     params.promptForEstimate ?? params.followupRun.prompt,
   );
+  const maxActiveTranscriptBytes = resolveMaxActiveTranscriptBytes(params.cfg);
+  const shouldCheckActiveTranscriptBytes = typeof maxActiveTranscriptBytes === "number";
   const transcriptUsageTokens =
     typeof freshPersistedTokens === "number"
       ? undefined
@@ -773,6 +759,26 @@ export async function runPreflightCompactionIfNeeded(params: {
           sessionFile: entry.sessionFile ?? params.followupRun.run.sessionFile,
           storePath: params.storePath,
         });
+  const transcriptSizeSnapshot =
+    shouldCheckActiveTranscriptBytes && transcriptUsageTokens?.transcriptByteSize === undefined
+      ? await readSessionLogSnapshot({
+          sessionId: entry.sessionId,
+          sessionEntry:
+            entry.sessionFile || !params.followupRun.run.sessionFile
+              ? entry
+              : { ...entry, sessionFile: params.followupRun.run.sessionFile },
+          sessionKey: params.sessionKey ?? params.followupRun.run.sessionKey,
+          opts: { storePath: params.storePath },
+          includeByteSize: true,
+          includeUsage: false,
+        })
+      : undefined;
+  const activeTranscriptBytes =
+    transcriptUsageTokens?.transcriptByteSize ?? transcriptSizeSnapshot?.byteSize;
+  const shouldCompactByTranscriptBytes =
+    typeof activeTranscriptBytes === "number" &&
+    typeof maxActiveTranscriptBytes === "number" &&
+    activeTranscriptBytes >= maxActiveTranscriptBytes;
   const stalePersistedPromptTokens = hasPersistedTotalTokens
     ? Math.floor(persistedTotalTokens)
     : undefined;
