fix: preserve codex preflight compaction route

amknight · amknight · commit 66354a42582b · 2026-05-25T12:21:54.000+10:00
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -11,6 +11,7 @@ Docs: https://docs.openclaw.ai
 - Scripts: remove stale Knip unused-file allowlist entries so the dead-code gate fails only on current findings.
 - Tests: normalize bundled plugin lifecycle probe paths and state-root lookup so native Windows release sweeps accept valid packaged plugin installs.
 - Config: keep benign legacy metadata write anomalies out of default doctor and config command output while preserving explicit anomaly logging for diagnostics.
+- Agents/Codex: route budget preflight compaction through the persisted Codex session model so Slack threads do not require separate plain OpenAI auth. Thanks @amknight.
 - Codex: log when implicit app-server `never` approvals are promoted for OpenClaw tool policy, including whether the trigger was a `before_tool_call` hook or trusted tool policy.
 - Google Vertex: support production ADC modes such as Workload Identity Federation, service-account credentials, and metadata-server ADC for the native Vertex transport. (#83971) Thanks @damianFelixPago.
 - Telegram: route normal `[telegram][diag]` polling diagnostics through `runtime.log` while keeping non-diag warnings and persistence failures on `runtime.error`, so healthy polling startup no longer looks like an error. Fixes #82957. (#82958) Thanks @galiniliev.
diff --git a/src/auto-reply/reply/agent-runner-memory.test.ts b/src/auto-reply/reply/agent-runner-memory.test.ts
@@ -78,6 +78,10 @@ type CompactEmbeddedPiSessionParams = {
   sessionFile?: string;
   sessionId?: string;
   trigger?: string;
+  provider?: string;
+  model?: string;
+  authProfileId?: string;
+  agentHarnessId?: string;
 };
 
 function requireRefreshQueuedFollowupSessionCall(index = 0) {
@@ -865,6 +869,59 @@ describe("runMemoryFlushIfNeeded", () => {
     expect(compactCall.currentTokenCount).toBe(347_000);
   });
 
+  it("uses the persisted Codex model route for OpenAI preflight compaction", async () => {
+    registerMemoryFlushPlanResolverForTest(() => ({
+      softThresholdTokens: 4_000,
+      forceFlushTranscriptBytes: 1_000_000_000,
+      reserveTokensFloor: 0,
+      prompt: "Pre-compaction memory flush.\nNO_REPLY",
+      systemPrompt: "Write memory to memory/YYYY-MM-DD.md.",
+      relativePath: "memory/2023-11-14.md",
+    }));
+    const sessionEntry: SessionEntry = {
+      sessionId: "session",
+      updatedAt: Date.now(),
+      totalTokens: 250_000,
+      totalTokensFresh: true,
+      modelProvider: "codex",
+      model: "gpt-5.5",
+      agentHarnessId: "codex",
+      authProfileOverride: "codex:work",
+    };
+
+    await runPreflightCompactionIfNeeded({
+      cfg: {
+        models: {
+          providers: {
+            openai: { models: [{ id: "gpt-5.5", contextWindow: 100_000 }] },
+            codex: { models: [{ id: "gpt-5.5", contextWindow: 100_000 }] },
+          },
+        },
+        agents: { defaults: { compaction: { memoryFlush: {} } } },
+      } as never,
+      followupRun: createTestFollowupRun({
+        provider: "openai",
+        model: "gpt-5.5",
+        sessionId: "session",
+        sessionKey: "main",
+      }),
+      defaultModel: "gpt-5.5",
+      sessionEntry,
+      sessionStore: { main: sessionEntry },
+      sessionKey: "main",
+      storePath: path.join(rootDir, "sessions.json"),
+      isHeartbeat: false,
+      replyOperation: createReplyOperation(),
+    });
+
+    expect(compactEmbeddedPiSessionMock).toHaveBeenCalledTimes(1);
+    const compactCall = requireCompactEmbeddedPiSessionCall();
+    expect(compactCall.provider).toBe("codex");
+    expect(compactCall.model).toBe("gpt-5.5");
+    expect(compactCall.agentHarnessId).toBe("codex");
+    expect(compactCall.authProfileId).toBe("codex:work");
+  });
+
   it("still compacts when a fresh persisted token total is over the threshold", async () => {
     registerMemoryFlushPlanResolverForTest(() => ({
       softThresholdTokens: 4_000,
diff --git a/src/auto-reply/reply/agent-runner-memory.ts b/src/auto-reply/reply/agent-runner-memory.ts
@@ -265,6 +265,25 @@ function resolveFollowupContextConfigProvider(params: {
   });
 }
 
+function resolvePreflightCompactionRunTarget(params: {
+  followupRun: FollowupRun;
+  sessionEntry?: SessionEntry;
+}): { provider: string; model: string; authProfileId?: string; agentHarnessId?: string } {
+  const run = params.followupRun.run;
+  const matchingSessionEntry =
+    params.sessionEntry?.sessionId === run.sessionId ? params.sessionEntry : undefined;
+  const authProfileId =
+    normalizeOptionalString(matchingSessionEntry?.authProfileOverride) ??
+    normalizeOptionalString(run.authProfileId);
+  const agentHarnessId = normalizeOptionalString(matchingSessionEntry?.agentHarnessId);
+  return {
+    provider: normalizeOptionalString(matchingSessionEntry?.modelProvider) ?? run.provider,
+    model: normalizeOptionalString(matchingSessionEntry?.model) ?? run.model,
+    ...(authProfileId ? { authProfileId } : {}),
+    ...(agentHarnessId ? { agentHarnessId } : {}),
+  };
+}
+
 function resolveVisibleMemoryFlushErrorPayloads(payloads?: ReplyPayload[]): ReplyPayload[] {
   return (payloads ?? []).filter(
     (payload) => payload.isError === true && isRenderablePayload(payload),
@@ -735,6 +754,10 @@ export async function runPreflightCompactionIfNeeded(params: {
     params.sessionKey ?? params.followupRun.run.sessionKey,
     { storePath: params.storePath },
   );
+  const compactionRunTarget = resolvePreflightCompactionRunTarget({
+    followupRun: params.followupRun,
+    sessionEntry: entry,
+  });
   const result = await memoryDeps.compactEmbeddedPiSession({
     sessionId: entry.sessionId,
     sessionKey: params.sessionKey,
@@ -753,10 +776,14 @@ export async function runPreflightCompactionIfNeeded(params: {
     agentDir: params.followupRun.run.agentDir,
     config: params.cfg,
     skillsSnapshot: entry.skillsSnapshot ?? params.followupRun.run.skillsSnapshot,
-    provider: params.followupRun.run.provider,
-    model: params.followupRun.run.model,
-    agentHarnessId:
-      entry.sessionId === params.followupRun.run.sessionId ? entry.agentHarnessId : undefined,
+    provider: compactionRunTarget.provider,
+    model: compactionRunTarget.model,
+    ...(compactionRunTarget.authProfileId
+      ? { authProfileId: compactionRunTarget.authProfileId }
+      : {}),
+    ...(compactionRunTarget.agentHarnessId
+      ? { agentHarnessId: compactionRunTarget.agentHarnessId }
+      : {}),
     thinkLevel: params.followupRun.run.thinkLevel,
     bashElevated: params.followupRun.run.bashElevated,
     trigger: "budget",
diff --git a/src/auto-reply/reply/agent-runner.runreplyagent.e2e.test.ts b/src/auto-reply/reply/agent-runner.runreplyagent.e2e.test.ts
@@ -413,6 +413,82 @@ describe("runReplyAgent heartbeat followup guard", () => {
   });
 });
 
+describe("runReplyAgent preflight compaction", () => {
+  it("keeps Codex-backed Slack sessions on the persisted Codex compaction route", async () => {
+    const dir = await mkdtemp(join(tmpdir(), "openclaw-codex-preflight-"));
+    try {
+      const sessionFile = join(dir, "session.jsonl");
+      const storePath = join(dir, "sessions.json");
+      await writeFile(
+        sessionFile,
+        `${JSON.stringify({ message: { role: "user", content: "hi" } })}\n`,
+        "utf8",
+      );
+      const sessionEntry: SessionEntry = {
+        sessionId: "session",
+        sessionFile,
+        updatedAt: Date.now(),
+        totalTokens: 250_000,
+        totalTokensFresh: true,
+        modelProvider: "codex",
+        model: "gpt-5.5",
+        agentHarnessId: "codex",
+      };
+      const sessionStore = { main: sessionEntry };
+      await writeFile(storePath, JSON.stringify(sessionStore), "utf8");
+      state.compactEmbeddedPiSessionMock.mockResolvedValueOnce({
+        ok: true,
+        compacted: true,
+        result: { tokensAfter: 42 },
+      });
+
+      const { run } = createMinimalRun({
+        sessionEntry,
+        sessionStore,
+        sessionKey: "main",
+        storePath,
+        sessionCtx: { Provider: "slack" },
+        runOverrides: {
+          provider: "openai",
+          model: "gpt-5.5",
+          sessionId: "session",
+          sessionFile,
+          messageProvider: "slack",
+          config: {
+            models: {
+              providers: {
+                openai: { models: [{ id: "gpt-5.5", contextWindow: 100_000 }] },
+                codex: { models: [{ id: "gpt-5.5", contextWindow: 100_000 }] },
+              },
+            },
+            agents: { defaults: { compaction: { memoryFlush: {} } } },
+          },
+        },
+      });
+
+      await run();
+
+      expect(state.compactEmbeddedPiSessionMock).toHaveBeenCalledTimes(1);
+      const compactCall = requireRecord(
+        mockCallArgs(state.compactEmbeddedPiSessionMock, "preflight compaction")[0],
+        "preflight compaction call",
+      );
+      expect(compactCall.provider).toBe("codex");
+      expect(compactCall.model).toBe("gpt-5.5");
+      expect(compactCall.agentHarnessId).toBe("codex");
+
+      const runCall = requireRecord(
+        mockCallArgs(state.runEmbeddedPiAgentMock, "run embedded pi agent")[0],
+        "embedded run call",
+      );
+      expect(runCall.provider).toBe("openai");
+      expect(runCall.model).toBe("gpt-5.5");
+    } finally {
+      await rm(dir, { recursive: true, force: true });
+    }
+  });
+});
+
 describe("runReplyAgent pending final delivery capture", () => {
   async function createSessionStoreFile(entry: SessionEntry) {
     const dir = await mkdtemp(join(tmpdir(), "openclaw-agent-runner-pending-"));
