@ouro.bot/cli 0.1.0-alpha.579 → 0.1.0-alpha.580

package/changelog.json

@@ -1,6 +1,14 @@
 {
   "_note": "This changelog is maintained as part of the PR/version-bump workflow. Agent-curated, not auto-generated. Agents read this file directly via read_file to understand what changed between versions.",
   "versions": [
+    {
+      "version": "0.1.0-alpha.580",
+      "changes": [
+        "OpenAI Realtime phone voice now queues manual `response.create` requests while another response is active, preventing SIP calls from hitting `conversation_already_has_active_response` and losing conversational state during tool followups.",
+        "Voice Realtime bootstrap now exposes the native voice/base tool set immediately instead of only `voice_end_call`, so early phone turns can still use tools while the full friend/MCP tool refresh finishes.",
+        "BlueBubbles webhook handling now claims message GUIDs before repair/hydrate work starts and releases those claims if repair discovers an already-processed message, preventing duplicate webhook deliveries from racing into overlapping agent turns for the same message."
+      ]
+    },
     {
       "version": "0.1.0-alpha.579",
       "changes": [

package/dist/senses/bluebubbles/index.js

@@ -919,9 +919,18 @@ async function handleBlueBubblesNormalizedEvent(event, resolvedDeps, source, opt
     let ownsInFlightMessage = false;
     let releaseInFlightAfterTurnSettles = false;
     let activeTurnId = null;
+    const inFlightKey = event.kind === "message"
+        ? {
+            sessionKey: options.preclaimedInFlight?.sessionKey ?? event.chat.sessionKey,
+            messageGuid: options.preclaimedInFlight?.messageGuid ?? event.messageGuid,
+        }
+        : null;
     if (event.kind === "message") {
         if ((0, processed_log_1.hasProcessedBlueBubblesMessage)(agentName, event.chat.sessionKey, event.messageGuid)
             || (0, processed_log_1.hasProcessedBlueBubblesMessageGuid)(agentName, event.messageGuid)) {
+            if (inFlightKey && options.preclaimedInFlight) {
+                endBlueBubblesMessageInFlight(inFlightKey.sessionKey, inFlightKey.messageGuid);
+            }
             (0, runtime_1.emitNervesEvent)({
                 component: "senses",
                 event: "senses.bluebubbles_recovery_skip",
@@ -935,7 +944,10 @@ async function handleBlueBubblesNormalizedEvent(event, resolvedDeps, source, opt
             });
             return { handled: true, notifiedAgent: false, kind: event.kind, reason: "already_processed" };
         }
-        if (!beginBlueBubblesMessageInFlight(event.chat.sessionKey, event.messageGuid)) {
+        if (options.preclaimedInFlight) {
+            ownsInFlightMessage = true;
+        }
+        else if (!beginBlueBubblesMessageInFlight(inFlightKey.sessionKey, inFlightKey.messageGuid)) {
             (0, runtime_1.emitNervesEvent)({
                 component: "senses",
                 event: "senses.bluebubbles_recovery_skip",
@@ -1084,8 +1096,8 @@ async function handleBlueBubblesNormalizedEvent(event, resolvedDeps, source, opt
                 timeoutTimer = setTimeout(() => {
                     const reason = new BlueBubblesRecoveryTurnTimeoutError(timeoutMs);
                     recoveryTimedOut = true;
-                    if (liveWebhookTimeout && ownsInFlightMessage && event.kind === "message") {
-                        endBlueBubblesMessageInFlight(event.chat.sessionKey, event.messageGuid);
+                    if (liveWebhookTimeout && ownsInFlightMessage && inFlightKey) {
+                        endBlueBubblesMessageInFlight(inFlightKey.sessionKey, inFlightKey.messageGuid);
                         ownsInFlightMessage = false;
                     }
                     else {
@@ -1190,8 +1202,8 @@ async function handleBlueBubblesNormalizedEvent(event, resolvedDeps, source, opt
                     });
                 })
                     .finally(() => {
-                    if (releaseInFlightAfterTurnSettles && ownsInFlightMessage && event.kind === "message") {
-                        endBlueBubblesMessageInFlight(event.chat.sessionKey, event.messageGuid);
+                    if (releaseInFlightAfterTurnSettles && ownsInFlightMessage && inFlightKey) {
+                        endBlueBubblesMessageInFlight(inFlightKey.sessionKey, inFlightKey.messageGuid);
                     }
                 });
                 /* v8 ignore stop */
@@ -1272,8 +1284,8 @@ async function handleBlueBubblesNormalizedEvent(event, resolvedDeps, source, opt
     finally {
         if (activeTurnId)
             (0, active_turns_1.finishBlueBubblesActiveTurn)(agentName, activeTurnId);
-        if (ownsInFlightMessage && event.kind === "message" && !releaseInFlightAfterTurnSettles) {
-            endBlueBubblesMessageInFlight(event.chat.sessionKey, event.messageGuid);
+        if (ownsInFlightMessage && inFlightKey && !releaseInFlightAfterTurnSettles) {
+            endBlueBubblesMessageInFlight(inFlightKey.sessionKey, inFlightKey.messageGuid);
         }
     }
 }
@@ -1346,8 +1358,24 @@ async function handleBlueBubblesEvent(payload, deps = {}) {
         });
         return handleBlueBubblesNormalizedEvent(normalized, resolvedDeps, "webhook");
     }
-    const event = await client.repairEvent(normalized);
-    return handleBlueBubblesNormalizedEvent(event, resolvedDeps, "webhook");
+    let preclaimedInFlight = false;
+    let handedPreclaimToTurn = false;
+    if (normalized.kind === "message") {
+        beginBlueBubblesMessageInFlight(normalized.chat.sessionKey, normalized.messageGuid);
+        preclaimedInFlight = true;
+    }
+    try {
+        const event = await client.repairEvent(normalized);
+        handedPreclaimToTurn = preclaimedInFlight && event.kind === "message";
+        return handleBlueBubblesNormalizedEvent(event, resolvedDeps, "webhook", handedPreclaimToTurn
+            ? { preclaimedInFlight: { sessionKey: normalized.chat.sessionKey, messageGuid: normalized.messageGuid } }
+            : {});
+    }
+    finally {
+        if (preclaimedInFlight && !handedPreclaimToTurn) {
+            endBlueBubblesMessageInFlight(normalized.chat.sessionKey, normalized.messageGuid);
+        }
+    }
 }
 function listPendingCapturedInboundMessages(agentName) {
     const seenMessageGuids = new Set();

package/dist/senses/voice/twilio-phone.js

@@ -1128,6 +1128,7 @@ const OPENAI_REALTIME_BARGE_IN_MIN_SPEECH_MS = 160;
 const OPENAI_REALTIME_BARGE_IN_RMS_THRESHOLD = 900;
 const OPENAI_REALTIME_MIN_VOICE_SPEED = 0.25;
 const OPENAI_REALTIME_MAX_VOICE_SPEED = 1.5;
+const OPENAI_REALTIME_RESPONSE_CREATE_GRACE_MS = 50;
 const OPENAI_SIP_OUTBOUND_AMD_GREETING_TIMEOUT_MS = 10_000;
 const OPENAI_SIP_UNSUPPORTED_TOOL_NAMES = new Set();
 const OPENAI_SIP_DEFAULT_API_BASE_URL = "https://api.openai.com/v1";
@@ -1435,18 +1436,7 @@ function realtimeBootstrapInstructions(agentName, voiceStyle) {
     ].filter(Boolean).join(" ");
 }
 function realtimeBootstrapTools() {
-    return [{
-            type: "function",
-            name: "voice_end_call",
-            description: "End the active live voice phone call after a natural goodbye.",
-            parameters: {
-                type: "object",
-                properties: {
-                    reason: { type: "string", description: "Short reason for ending the call." },
-                },
-                additionalProperties: false,
-            },
-        }];
+    return realtimeToolsFromChatTools((0, tools_1.getToolsForChannel)((0, channel_1.getChannelCapabilities)("voice")));
 }
 function timeoutAfter(ms) {
     return new Promise((resolve) => {
@@ -1507,6 +1497,10 @@ class TwilioOpenAIRealtimeMediaStreamSession {
     playbackMarks = new Map();
     toolResponses = new Map();
     completedRealtimeResponseIds = new Set();
+    activeRealtimeResponseId = null;
+    pendingRealtimeResponse = null;
+    pendingRealtimeResponseTimer = null;
+    responseCreateHoldUntilMs = 0;
     initialAudio;
     initialAudioPlayed = false;
     callerBargeInSpeechMs = 0;
@@ -1813,12 +1807,7 @@ class TwilioOpenAIRealtimeMediaStreamSession {
                 createdAt: new Date().toISOString(),
             }, { From: this.to, To: this.from })
             : callConnectedPrompt({ From: this.from, To: this.to });
-        this.sendOpenAI({
-            type: "response.create",
-            response: {
-                instructions: promptText,
-            },
-        });
+        this.requestRealtimeResponse({ instructions: promptText });
     }
     handleMedia(media) {
         const payload = stringField(media?.payload);
@@ -1873,6 +1862,10 @@ class TwilioOpenAIRealtimeMediaStreamSession {
             return;
         }
         const type = typeof event.type === "string" ? event.type : "";
+        if (type === "response.created") {
+            this.noteRealtimeResponseCreated(event);
+            return;
+        }
         if (type === "response.output_audio.delta" && typeof event.delta === "string") {
             this.handleOpenAIAudioDelta(event);
             return;
@@ -1894,7 +1887,9 @@ class TwilioOpenAIRealtimeMediaStreamSession {
             return;
         }
         if (type === "response.done") {
-            if (this.completeRealtimeToolResponse(realtimeResponseId(event)))
+            const responseId = realtimeResponseId(event);
+            this.noteRealtimeResponseDone(responseId);
+            if (this.completeRealtimeToolResponse(responseId))
                 return;
             void this.playInitialAudioAfterGreeting();
             this.completeHangupIfReady("response_done");
@@ -2009,7 +2004,7 @@ class TwilioOpenAIRealtimeMediaStreamSession {
         this.toolResponses.delete(responseId);
         if (state.suppressFollowup)
             return true;
-        this.sendOpenAI({ type: "response.create" });
+        this.requestRealtimeResponse();
         return true;
     }
     async runRealtimeTool(event) {
@@ -2058,8 +2053,62 @@ class TwilioOpenAIRealtimeMediaStreamSession {
             },
         });
         if (!this.completeRealtimeToolCall(responseId, callId) && !coordinated) {
-            this.sendOpenAI({ type: "response.create" });
+            this.requestRealtimeResponse();
+        }
+    }
+    noteRealtimeResponseCreated(event) {
+        const responseId = realtimeResponseId(event);
+        if (responseId)
+            this.activeRealtimeResponseId = responseId;
+    }
+    noteRealtimeResponseDone(responseId) {
+        if (!responseId || this.activeRealtimeResponseId === responseId) {
+            this.activeRealtimeResponseId = null;
+        }
+        this.responseCreateHoldUntilMs = Math.max(this.responseCreateHoldUntilMs, Date.now() + OPENAI_REALTIME_RESPONSE_CREATE_GRACE_MS);
+        this.schedulePendingRealtimeResponse(OPENAI_REALTIME_RESPONSE_CREATE_GRACE_MS);
+    }
+    requestRealtimeResponse(response) {
+        if (this.closed)
+            return;
+        const waitMs = Math.max(0, this.responseCreateHoldUntilMs - Date.now());
+        if (this.activeRealtimeResponseId || waitMs > 0) {
+            const pendingResponse = response ?? this.pendingRealtimeResponse?.response;
+            this.pendingRealtimeResponse = pendingResponse ? { response: pendingResponse } : {};
+            if (!this.activeRealtimeResponseId)
+                this.schedulePendingRealtimeResponse(waitMs);
+            return;
         }
+        this.sendRealtimeResponseCreate(response ? { response } : {});
+    }
+    schedulePendingRealtimeResponse(delayMs) {
+        if (!this.pendingRealtimeResponse)
+            return;
+        if (this.pendingRealtimeResponseTimer)
+            clearTimeout(this.pendingRealtimeResponseTimer);
+        this.pendingRealtimeResponseTimer = setTimeout(() => {
+            this.pendingRealtimeResponseTimer = null;
+            this.flushPendingRealtimeResponse();
+        }, Math.max(0, delayMs));
+        this.pendingRealtimeResponseTimer.unref?.();
+    }
+    flushPendingRealtimeResponse() {
+        if (!this.pendingRealtimeResponse || this.closed || this.activeRealtimeResponseId)
+            return;
+        const waitMs = Math.max(0, this.responseCreateHoldUntilMs - Date.now());
+        if (waitMs > 0) {
+            this.schedulePendingRealtimeResponse(waitMs);
+            return;
+        }
+        const pending = this.pendingRealtimeResponse;
+        this.pendingRealtimeResponse = null;
+        this.sendRealtimeResponseCreate(pending);
+    }
+    sendRealtimeResponseCreate(request) {
+        this.sendOpenAI({
+            type: "response.create",
+            ...(request.response ? { response: request.response } : {}),
+        });
     }
     flushPendingAudio() {
         const pending = this.pendingAudioPayloads.splice(0);
@@ -2176,6 +2225,10 @@ class TwilioOpenAIRealtimeMediaStreamSession {
         if (this.openaiWs && (this.openaiWs.readyState === ws_1.WebSocket.OPEN || this.openaiWs.readyState === ws_1.WebSocket.CONNECTING)) {
             this.openaiWs.close();
         }
+        if (this.pendingRealtimeResponseTimer) {
+            clearTimeout(this.pendingRealtimeResponseTimer);
+            this.pendingRealtimeResponseTimer = null;
+        }
         this.lifecycle?.onClose?.(this, { callSid: this.callSid, outboundId: this.outboundId });
         (0, runtime_1.emitNervesEvent)({
             component: "senses",
@@ -2207,6 +2260,10 @@ class OpenAISipPhoneSession {
     sessionMessages = [];
     toolResponses = new Map();
     completedRealtimeResponseIds = new Set();
+    activeRealtimeResponseId = null;
+    pendingRealtimeResponse = null;
+    pendingRealtimeResponseTimer = null;
+    responseCreateHoldUntilMs = 0;
     constructor(options, metadata, registry) {
         this.options = options;
         this.metadata = metadata;
@@ -2722,11 +2779,8 @@ class OpenAISipPhoneSession {
         if (!this.openaiWs || this.openaiWs.readyState !== ws_1.WebSocket.OPEN)
             return;
         this.initialGreetingSent = true;
-        this.sendOpenAI({
-            type: "response.create",
-            response: {
-                instructions: openAISipCallConnectedPrompt(this.metadata, this.options.openaiRealtime?.voiceStyle),
-            },
+        this.requestRealtimeResponse({
+            instructions: openAISipCallConnectedPrompt(this.metadata, this.options.openaiRealtime?.voiceStyle),
         });
     }
     handleOpenAIMessage(raw) {
@@ -2738,6 +2792,10 @@ class OpenAISipPhoneSession {
             return;
         }
         const type = typeof event.type === "string" ? event.type : "";
+        if (type === "response.created") {
+            this.noteRealtimeResponseCreated(event);
+            return;
+        }
         if (type === "conversation.item.input_audio_transcription.completed" && typeof event.transcript === "string") {
             this.recordOutboundAmdTranscriptCandidate(event.transcript);
             this.appendTranscript("user", event.transcript);
@@ -2752,7 +2810,9 @@ class OpenAISipPhoneSession {
             return;
         }
         if (type === "response.done") {
-            if (this.completeRealtimeToolResponse(realtimeResponseId(event)))
+            const responseId = realtimeResponseId(event);
+            this.noteRealtimeResponseDone(responseId);
+            if (this.completeRealtimeToolResponse(responseId))
                 return;
             this.completeHangupIfReady("response_done");
             return;
@@ -2811,7 +2871,7 @@ class OpenAISipPhoneSession {
             this.completeHangupIfReady("tool_response_done");
             return true;
         }
-        this.sendOpenAI({ type: "response.create" });
+        this.requestRealtimeResponse();
         return true;
     }
     async runRealtimeTool(event) {
@@ -2860,9 +2920,63 @@ class OpenAISipPhoneSession {
             },
         });
         if (!this.completeRealtimeToolCall(responseId, callId) && !coordinated) {
-            this.sendOpenAI({ type: "response.create" });
+            this.requestRealtimeResponse();
         }
     }
+    noteRealtimeResponseCreated(event) {
+        const responseId = realtimeResponseId(event);
+        if (responseId)
+            this.activeRealtimeResponseId = responseId;
+    }
+    noteRealtimeResponseDone(responseId) {
+        if (!responseId || this.activeRealtimeResponseId === responseId) {
+            this.activeRealtimeResponseId = null;
+        }
+        this.responseCreateHoldUntilMs = Math.max(this.responseCreateHoldUntilMs, Date.now() + OPENAI_REALTIME_RESPONSE_CREATE_GRACE_MS);
+        this.schedulePendingRealtimeResponse(OPENAI_REALTIME_RESPONSE_CREATE_GRACE_MS);
+    }
+    requestRealtimeResponse(response) {
+        if (this.closed)
+            return;
+        const waitMs = Math.max(0, this.responseCreateHoldUntilMs - Date.now());
+        if (this.activeRealtimeResponseId || waitMs > 0) {
+            const pendingResponse = response ?? this.pendingRealtimeResponse?.response;
+            this.pendingRealtimeResponse = pendingResponse ? { response: pendingResponse } : {};
+            if (!this.activeRealtimeResponseId)
+                this.schedulePendingRealtimeResponse(waitMs);
+            return;
+        }
+        this.sendRealtimeResponseCreate(response ? { response } : {});
+    }
+    schedulePendingRealtimeResponse(delayMs) {
+        if (!this.pendingRealtimeResponse)
+            return;
+        if (this.pendingRealtimeResponseTimer)
+            clearTimeout(this.pendingRealtimeResponseTimer);
+        this.pendingRealtimeResponseTimer = setTimeout(() => {
+            this.pendingRealtimeResponseTimer = null;
+            this.flushPendingRealtimeResponse();
+        }, Math.max(0, delayMs));
+        this.pendingRealtimeResponseTimer.unref?.();
+    }
+    flushPendingRealtimeResponse() {
+        if (!this.pendingRealtimeResponse || this.closed || this.activeRealtimeResponseId)
+            return;
+        const waitMs = Math.max(0, this.responseCreateHoldUntilMs - Date.now());
+        if (waitMs > 0) {
+            this.schedulePendingRealtimeResponse(waitMs);
+            return;
+        }
+        const pending = this.pendingRealtimeResponse;
+        this.pendingRealtimeResponse = null;
+        this.sendRealtimeResponseCreate(pending);
+    }
+    sendRealtimeResponseCreate(request) {
+        this.sendOpenAI({
+            type: "response.create",
+            ...(request.response ? { response: request.response } : {}),
+        });
+    }
     requestHangupFromTool() {
         if (this.closed)
             return;
@@ -2932,6 +3046,10 @@ class OpenAISipPhoneSession {
         if (this.openaiWs && (this.openaiWs.readyState === ws_1.WebSocket.OPEN || this.openaiWs.readyState === ws_1.WebSocket.CONNECTING)) {
             this.openaiWs.close();
         }
+        if (this.pendingRealtimeResponseTimer) {
+            clearTimeout(this.pendingRealtimeResponseTimer);
+            this.pendingRealtimeResponseTimer = null;
+        }
         (0, runtime_1.emitNervesEvent)({
             component: "senses",
             event: "senses.voice_openai_sip_call_stop",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ouro.bot/cli",
-  "version": "0.1.0-alpha.579",
+  "version": "0.1.0-alpha.580",
   "main": "dist/heart/daemon/ouro-entry.js",
   "bin": {
     "cli": "dist/heart/daemon/ouro-bot-entry.js",