From 3efd1542d382b3e175945c9a01338d9332fce648 Mon Sep 17 00:00:00 2001
From: Roger Chappel <miscanalysis@gmail.com>
Date: Thu, 21 May 2026 17:18:04 +1000
Subject: [PATCH] fix: persist realtime consult results

---
 .../[id]/talk/realtime/relay/route.test.ts    |  1 +
 .../[id]/talk/realtime/relay/route.ts         |  2 +-
 src/lib/realtime-voice-client.ts              | 18 +++++++-
 src/lib/realtime-voice-gateway-relay.ts       | 44 +++++++++++++------
 4 files changed, 48 insertions(+), 17 deletions(-)

diff --git a/src/app/api/runtimes/[id]/talk/realtime/relay/route.test.ts b/src/app/api/runtimes/[id]/talk/realtime/relay/route.test.ts
index 9a45309..cfff020 100644
--- a/src/app/api/runtimes/[id]/talk/realtime/relay/route.test.ts
+++ b/src/app/api/runtimes/[id]/talk/realtime/relay/route.test.ts
@@ -161,6 +161,7 @@ describe("POST /api/runtimes/[id]/talk/realtime/relay", () => {
       result: {
         delegated: true,
         runId: "run_1",
+        finalText: "The repo is a CrewCMD app.",
         result: { ok: true },
       },
     });
diff --git a/src/app/api/runtimes/[id]/talk/realtime/relay/route.ts b/src/app/api/runtimes/[id]/talk/realtime/relay/route.ts
index 8539052..d03d3bd 100644
--- a/src/app/api/runtimes/[id]/talk/realtime/relay/route.ts
+++ b/src/app/api/runtimes/[id]/talk/realtime/relay/route.ts
@@ -137,7 +137,7 @@ async function runRealtimeToolCall(
         callId: params.callId,
         result: { result: text },
       });
-      return { delegated: true, runId, result };
+      return { delegated: true, runId, result, finalText: text };
     } catch (error) {
       const message = error instanceof Error ? error.message : "OpenClaw realtime tool call failed";
       await client.realtimeRelayToolResult({
diff --git a/src/lib/realtime-voice-client.ts b/src/lib/realtime-voice-client.ts
index b348055..e9f2715 100644
--- a/src/lib/realtime-voice-client.ts
+++ b/src/lib/realtime-voice-client.ts
@@ -46,6 +46,13 @@ export interface RealtimeRelayToolCall {
   args?: unknown;
 }
 
+export interface RealtimeRelayToolCallResult {
+  delegated?: boolean;
+  runId?: string;
+  finalText?: string;
+  result?: unknown;
+}
+
 export async function startRealtimeVoiceSession(
   request: RealtimeVoiceSessionRequest,
 ): Promise<RealtimeVoiceSession> {
@@ -114,11 +121,17 @@ export async function sendRealtimeRelayToolResult(
   });
 }
 
-export async function sendRealtimeRelayToolCall(runtimeId: string, toolCall: RealtimeRelayToolCall): Promise<void> {
-  await postRealtimeRelay(runtimeId, {
+export async function sendRealtimeRelayToolCall(
+  runtimeId: string,
+  toolCall: RealtimeRelayToolCall,
+): Promise<RealtimeRelayToolCallResult> {
+  const data = await postRealtimeRelay(runtimeId, {
     action: "toolCall",
     ...toolCall,
   });
+  return data.result && typeof data.result === "object"
+    ? data.result as RealtimeRelayToolCallResult
+    : {};
 }
 
 export async function stopRealtimeRelay(runtimeId: string, relaySessionId: string): Promise<void> {
@@ -145,6 +158,7 @@ async function postRealtimeRelay(runtimeId: string, body: Record<string, unknown
   if (!response.ok) {
     throw new Error(await readRealtimeVoiceError(response, "Realtime relay request failed"));
   }
+  return await response.json().catch(() => ({})) as { result?: unknown };
 }
 
 async function readRealtimeVoiceError(response: Response, fallback: string) {
diff --git a/src/lib/realtime-voice-gateway-relay.ts b/src/lib/realtime-voice-gateway-relay.ts
index ab4b563..59a5239 100644
--- a/src/lib/realtime-voice-gateway-relay.ts
+++ b/src/lib/realtime-voice-gateway-relay.ts
@@ -97,6 +97,7 @@ export class RealtimeGatewayRelaySession {
   private cancelRequestedForPlayback = false;
   private speechFramesDuringPlayback = 0;
   private outputStartedAtMs: number | null = null;
+  private pendingToolCalls = 0;
   private readonly bargeInProfile = resolveRealtimeBargeInProfile();
 
   constructor(
@@ -229,6 +230,7 @@ export class RealtimeGatewayRelaySession {
         return;
       case "transcript":
         if (event.role && event.text) {
+          if (event.role === "assistant" && this.pendingToolCalls > 0) return;
           this.callbacks.onTranscript?.({
             role: event.role,
             text: event.text,
@@ -299,20 +301,34 @@ export class RealtimeGatewayRelaySession {
     const name = event.name;
 
     this.callbacks.onStatus?.("processing", "Consulting OpenClaw");
-    void sendRealtimeRelayToolCall(this.runtimeId, {
-      relaySessionId,
-      sessionKey,
-      callId,
-      name,
-      args: event.args ?? {},
-    }).catch((error) => {
-      const message = error instanceof Error ? error.message : String(error);
-      this.callbacks.onError?.(message);
-      void sendRealtimeRelayToolResult(this.runtimeId, relaySessionId, callId, {
-        error: message,
-        name,
-      }).catch(() => {});
-    });
+    this.pendingToolCalls += 1;
+    void (async () => {
+      try {
+        const result = await sendRealtimeRelayToolCall(this.runtimeId, {
+          relaySessionId,
+          sessionKey,
+          callId,
+          name,
+          args: event.args ?? {},
+        });
+        if (result.finalText?.trim()) {
+          this.callbacks.onTranscript?.({
+            role: "assistant",
+            text: result.finalText.trim(),
+            final: true,
+          });
+        }
+      } catch (error) {
+        const message = error instanceof Error ? error.message : String(error);
+        this.callbacks.onError?.(message);
+        void sendRealtimeRelayToolResult(this.runtimeId, relaySessionId, callId, {
+          error: message,
+          name,
+        }).catch(() => {});
+      } finally {
+        this.pendingToolCalls = Math.max(0, this.pendingToolCalls - 1);
+      }
+    })();
   }
 
   private cancelOutputForBargeIn(): void {