Remove Twilio dependencies and cleanup references

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

backend/config.py

@@ -15,12 +15,6 @@ class Settings(BaseSettings):
     openrouter_api_key: str = os.getenv("OPENROUTER_API_KEY", "")
     inworld_api_key: str = os.getenv("INWORLD_API_KEY", "")
 
-    # Twilio Settings
-    twilio_account_sid: str = os.getenv("TWILIO_ACCOUNT_SID", "")
-    twilio_auth_token: str = os.getenv("TWILIO_AUTH_TOKEN", "")
-    twilio_phone_number: str = os.getenv("TWILIO_PHONE_NUMBER", "")
-    twilio_webhook_base_url: str = os.getenv("TWILIO_WEBHOOK_BASE_URL", "")
-
     # LLM Settings
     llm_provider: str = "openrouter"  # "openrouter" or "ollama"
     openrouter_model: str = "anthropic/claude-3-haiku"

backend/main.py

@@ -4,20 +4,17 @@ import uuid
 import asyncio
 from dataclasses import dataclass, field
 from pathlib import Path
-from fastapi import FastAPI, HTTPException, UploadFile, File, Form, WebSocket, WebSocketDisconnect
+from fastapi import FastAPI, HTTPException, WebSocket, WebSocketDisconnect
 from fastapi.staticfiles import StaticFiles
-from fastapi.responses import FileResponse, Response
-from twilio.twiml.voice_response import VoiceResponse
+from fastapi.responses import FileResponse
 import json
-import base64
-import audioop
 import time
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel
 from typing import Optional
 
 from .config import settings
-from .services.twilio_service import TwilioService
+from .services.caller_service import CallerService
 from .services.transcription import transcribe_audio
 from .services.llm import llm_service
 from .services.tts import generate_speech
@@ -247,7 +244,7 @@ def generate_caller_background(base: dict) -> str:
 
     return f"""{age}, {job} {location}. {problem.capitalize()}. {interest1.capitalize()}, {interest2}. {quirk1.capitalize()}, {quirk2}."""
 
-def get_caller_prompt(caller: dict, conversation_summary: str = "") -> str:
+def get_caller_prompt(caller: dict, conversation_summary: str = "", show_history: str = "") -> str:
     """Generate a natural system prompt for a caller"""
     context = ""
     if conversation_summary:
@@ -257,10 +254,14 @@ CONVERSATION SO FAR:
 Continue naturally. Don't repeat yourself.
 """
 
+    history = ""
+    if show_history:
+        history = f"\n{show_history}\n"
+
     return f"""You're {caller['name']}, calling a late-night radio show. You trust this host.
 
 {caller['vibe']}
-{context}
+{history}{context}
 HOW TO TALK:
 - Sound like a real person chatting, not writing.
 - Keep responses to 2-3 sentences. Enough to make your point, short enough for back-and-forth.
@@ -401,7 +402,7 @@ class Session:
 
 
 session = Session()
-twilio_service = TwilioService()
+caller_service = CallerService()
 
 
 # --- Static Files ---
@@ -415,6 +416,11 @@ async def index():
     return FileResponse(frontend_dir / "index.html")
 
 
+@app.get("/call-in")
+async def call_in_page():
+    return FileResponse(frontend_dir / "call-in.html")
+
+
 # --- Request Models ---
 
 class ChatRequest(BaseModel):
@@ -611,9 +617,10 @@ async def chat(request: ChatRequest):
 
     session.add_message("user", request.text)
 
-    # Include conversation summary for context
+    # Include conversation summary and show history for context
     conversation_summary = session.get_conversation_summary()
-    system_prompt = get_caller_prompt(session.caller, conversation_summary)
+    show_history = session.get_show_history()
+    system_prompt = get_caller_prompt(session.caller, conversation_summary, show_history)
 
     response = await llm_service.generate(
         messages=session.conversation[-10:],  # Reduced history for speed
@@ -665,9 +672,9 @@ async def text_to_speech(request: TTSRequest):
 
     # Also send to active real callers so they hear the AI
     if session.active_real_caller:
-        call_sid = session.active_real_caller["call_sid"]
+        caller_id = session.active_real_caller["caller_id"]
         asyncio.create_task(
-            twilio_service.send_audio_to_caller(call_sid, audio_bytes, 24000)
+            caller_service.send_audio_to_caller(caller_id, audio_bytes, 24000)
         )
 
     return {"status": "playing", "duration": len(audio_bytes) / 2 / 24000}
@@ -775,67 +782,143 @@ async def update_settings(data: dict):
     return llm_service.get_settings()
 
 
-# --- Twilio Webhook & Queue Endpoints ---
+# --- Browser Caller WebSocket ---
 
-@app.post("/api/twilio/voice")
-async def twilio_voice_webhook(
-    CallSid: str = Form(...),
-    From: str = Form(...),
-):
-    """Handle incoming Twilio call — greet and enqueue"""
-    twilio_service.add_to_queue(CallSid, From)
+@app.websocket("/api/caller/stream")
+async def caller_audio_stream(websocket: WebSocket):
+    """Handle browser caller WebSocket — bidirectional audio"""
+    await websocket.accept()
 
-    response = VoiceResponse()
-    response.say("You're calling Luke at the Roost. Hold tight, we'll get to you.", voice="alice")
-    response.enqueue(
-        "radio_show",
-        wait_url="/api/twilio/hold-music",
-        wait_url_method="POST",
-    )
-    return Response(content=str(response), media_type="application/xml")
+    caller_id = str(uuid.uuid4())[:8]
+    caller_name = "Anonymous"
+    audio_buffer = bytearray()
+    CHUNK_DURATION_S = 3
+    SAMPLE_RATE = 16000
+    chunk_samples = CHUNK_DURATION_S * SAMPLE_RATE
+
+    try:
+        # Wait for join message
+        join_data = await websocket.receive_text()
+        join_msg = json.loads(join_data)
+        if join_msg.get("type") == "join":
+            caller_name = join_msg.get("name", "Anonymous").strip() or "Anonymous"
+
+        # Add to queue
+        caller_service.add_to_queue(caller_id, caller_name)
+        caller_service.register_websocket(caller_id, websocket)
+
+        # Notify caller they're queued
+        queue = caller_service.get_queue()
+        position = next((i+1 for i, c in enumerate(queue) if c["caller_id"] == caller_id), 0)
+        await websocket.send_text(json.dumps({
+            "status": "queued",
+            "caller_id": caller_id,
+            "position": position,
+        }))
+
+        # Main loop — handle both text and binary messages
+        while True:
+            message = await websocket.receive()
+
+            if message.get("type") == "websocket.disconnect":
+                break
+
+            if "bytes" in message and message["bytes"]:
+                # Binary audio data — only process if caller is on air
+                call_info = caller_service.active_calls.get(caller_id)
+                if not call_info:
+                    continue  # Still in queue, ignore audio
+
+                pcm_data = message["bytes"]
+                audio_buffer.extend(pcm_data)
+
+                # Route to Loopback channel
+                channel = call_info["channel"]
+                audio_service.route_real_caller_audio(pcm_data, channel, SAMPLE_RATE)
+
+                # Transcribe when we have enough audio
+                if len(audio_buffer) >= chunk_samples * 2:
+                    pcm_chunk = bytes(audio_buffer[:chunk_samples * 2])
+                    audio_buffer = audio_buffer[chunk_samples * 2:]
+                    asyncio.create_task(
+                        _handle_real_caller_transcription(caller_id, pcm_chunk, SAMPLE_RATE)
+                    )
+
+            elif "text" in message and message["text"]:
+                # Control messages (future use)
+                pass
+
+    except WebSocketDisconnect:
+        print(f"[Caller WS] Disconnected: {caller_id} ({caller_name})")
+    except Exception as e:
+        print(f"[Caller WS] Error: {e}")
+    finally:
+        caller_service.unregister_websocket(caller_id)
+        # If still in queue, remove
+        caller_service.remove_from_queue(caller_id)
+        # If on air, clean up
+        if caller_id in caller_service.active_calls:
+            caller_service.hangup(caller_id)
+            if session.active_real_caller and session.active_real_caller.get("caller_id") == caller_id:
+                session.active_real_caller = None
+        # Transcribe remaining audio
+        if audio_buffer:
+            asyncio.create_task(
+                _handle_real_caller_transcription(caller_id, bytes(audio_buffer), SAMPLE_RATE)
+            )
 
 
-@app.post("/api/twilio/hold-music")
-async def twilio_hold_music():
-    """Serve hold music TwiML for queued callers"""
-    response = VoiceResponse()
-    response.say("Please hold, you'll be on air shortly.", voice="alice")
-    response.pause(length=30)
-    return Response(content=str(response), media_type="application/xml")
+# --- Host Audio Broadcast ---
 
+async def _broadcast_host_audio(pcm_bytes: bytes):
+    """Send host mic audio to all active real callers"""
+    for caller_id in list(caller_service.active_calls.keys()):
+        try:
+            await caller_service.send_audio_to_caller(caller_id, pcm_bytes, 16000)
+        except Exception:
+            pass
+
+
+def _host_audio_sync_callback(pcm_bytes: bytes):
+    """Sync wrapper to schedule async broadcast from audio thread"""
+    try:
+        loop = asyncio.get_event_loop()
+        if loop.is_running():
+            loop.call_soon_threadsafe(
+                asyncio.ensure_future, _broadcast_host_audio(pcm_bytes)
+            )
+    except Exception:
+        pass
+
+
+# --- Queue Endpoints ---
 
 @app.get("/api/queue")
 async def get_call_queue():
     """Get list of callers waiting in queue"""
-    return {"queue": twilio_service.get_queue()}
+    return {"queue": caller_service.get_queue()}
 
 
-@app.post("/api/queue/take/{call_sid}")
-async def take_call_from_queue(call_sid: str):
+@app.post("/api/queue/take/{caller_id}")
+async def take_call_from_queue(caller_id: str):
     """Take a caller off hold and put them on air"""
     try:
-        call_info = twilio_service.take_call(call_sid)
+        call_info = caller_service.take_call(caller_id)
     except ValueError as e:
         raise HTTPException(404, str(e))
 
     session.active_real_caller = {
-        "call_sid": call_info["call_sid"],
-        "phone": call_info["phone"],
+        "caller_id": call_info["caller_id"],
         "channel": call_info["channel"],
         "name": call_info["name"],
     }
 
-    # Connect Twilio media stream by updating the call
-    from twilio.rest import Client as TwilioClient
-    if settings.twilio_account_sid and settings.twilio_auth_token:
-        client = TwilioClient(settings.twilio_account_sid, settings.twilio_auth_token)
-        twiml = VoiceResponse()
-        connect = twiml.connect()
-        connect.stream(
-            url=f"wss://{settings.twilio_webhook_base_url.replace('https://', '')}/api/twilio/stream",
-            name=call_sid,
-        )
-        client.calls(call_sid).update(twiml=str(twiml))
+    # Notify caller they're on air via WebSocket
+    await caller_service.notify_caller(caller_id, {"status": "on_air", "channel": call_info["channel"]})
+
+    # Start host mic streaming if this is the first real caller
+    if len(caller_service.active_calls) == 1:
+        audio_service.start_host_stream(_host_audio_sync_callback)
 
     return {
         "status": "on_air",
@@ -843,97 +926,17 @@ async def take_call_from_queue(call_sid: str):
     }
 
 
-@app.post("/api/queue/drop/{call_sid}")
-async def drop_from_queue(call_sid: str):
+@app.post("/api/queue/drop/{caller_id}")
+async def drop_from_queue(caller_id: str):
     """Drop a caller from the queue"""
-    twilio_service.remove_from_queue(call_sid)
-
-    # Hang up the Twilio call
-    from twilio.rest import Client as TwilioClient
-    if settings.twilio_account_sid and settings.twilio_auth_token:
-        try:
-            client = TwilioClient(settings.twilio_account_sid, settings.twilio_auth_token)
-            client.calls(call_sid).update(status="completed")
-        except Exception as e:
-            print(f"[Twilio] Failed to end call {call_sid}: {e}")
-
+    caller_service.remove_from_queue(caller_id)
+    await caller_service.disconnect_caller(caller_id)
     return {"status": "dropped"}
 
 
-# --- Twilio WebSocket Media Stream ---
-
-@app.websocket("/api/twilio/stream")
-async def twilio_media_stream(websocket: WebSocket):
-    """Handle Twilio Media Streams WebSocket — bidirectional audio"""
-    await websocket.accept()
-    print("[Twilio WS] Media stream connected")
-
-    call_sid = None
-    stream_sid = None
-    audio_buffer = bytearray()
-    CHUNK_DURATION_S = 3  # Transcribe every 3 seconds of audio
-    MULAW_SAMPLE_RATE = 8000
-    chunk_samples = CHUNK_DURATION_S * MULAW_SAMPLE_RATE
-
-    try:
-        while True:
-            data = await websocket.receive_text()
-            msg = json.loads(data)
-            event = msg.get("event")
-
-            if event == "start":
-                stream_sid = msg["start"]["streamSid"]
-                call_sid = msg["start"]["callSid"]
-                twilio_service.register_websocket(call_sid, websocket)
-                if call_sid in twilio_service.active_calls:
-                    twilio_service.active_calls[call_sid]["stream_sid"] = stream_sid
-                print(f"[Twilio WS] Stream started: {stream_sid} for call {call_sid}")
-
-            elif event == "media":
-                # Decode mulaw audio from base64
-                payload = base64.b64decode(msg["media"]["payload"])
-                # Convert mulaw to 16-bit PCM
-                pcm_data = audioop.ulaw2lin(payload, 2)
-                audio_buffer.extend(pcm_data)
-
-                # Get channel for this caller
-                call_info = twilio_service.active_calls.get(call_sid)
-                if call_info:
-                    channel = call_info["channel"]
-                    # Route PCM to the caller's dedicated Loopback channel
-                    audio_service.route_real_caller_audio(pcm_data, channel, MULAW_SAMPLE_RATE)
-
-                # When we have enough audio, transcribe
-                if len(audio_buffer) >= chunk_samples * 2:  # 2 bytes per sample
-                    pcm_chunk = bytes(audio_buffer[:chunk_samples * 2])
-                    audio_buffer = audio_buffer[chunk_samples * 2:]
-
-                    # Transcribe in background
-                    asyncio.create_task(
-                        _handle_real_caller_transcription(call_sid, pcm_chunk, MULAW_SAMPLE_RATE)
-                    )
-
-            elif event == "stop":
-                print(f"[Twilio WS] Stream stopped: {stream_sid}")
-                break
-
-    except WebSocketDisconnect:
-        print(f"[Twilio WS] Disconnected: {call_sid}")
-    except Exception as e:
-        print(f"[Twilio WS] Error: {e}")
-    finally:
-        if call_sid:
-            twilio_service.unregister_websocket(call_sid)
-        # Transcribe any remaining audio
-        if audio_buffer and call_sid:
-            asyncio.create_task(
-                _handle_real_caller_transcription(call_sid, bytes(audio_buffer), MULAW_SAMPLE_RATE)
-            )
-
-
-async def _handle_real_caller_transcription(call_sid: str, pcm_data: bytes, sample_rate: int):
+async def _handle_real_caller_transcription(caller_id: str, pcm_data: bytes, sample_rate: int):
     """Transcribe a chunk of real caller audio and add to conversation"""
-    call_info = twilio_service.active_calls.get(call_sid)
+    call_info = caller_service.active_calls.get(caller_id)
     if not call_info:
         return
 
@@ -979,7 +982,8 @@ async def _check_ai_auto_respond(real_caller_text: str, real_caller_name: str):
 
     # Generate full response
     conversation_summary = session.get_conversation_summary()
-    system_prompt = get_caller_prompt(session.caller, conversation_summary)
+    show_history = session.get_show_history()
+    system_prompt = get_caller_prompt(session.caller, conversation_summary, show_history)
 
     response = await llm_service.generate(
         messages=session.conversation[-10:],
@@ -1003,6 +1007,120 @@ async def _check_ai_auto_respond(real_caller_text: str, real_caller_name: str):
     thread.start()
 
 
+# --- Follow-Up & Session Control Endpoints ---
+
+@app.post("/api/hangup/real")
+async def hangup_real_caller():
+    """Hang up on real caller — summarize call and store in history"""
+    if not session.active_real_caller:
+        raise HTTPException(400, "No active real caller")
+
+    caller_id = session.active_real_caller["caller_id"]
+    caller_name = session.active_real_caller["name"]
+
+    # Summarize the conversation
+    summary = ""
+    if session.conversation:
+        transcript_text = "\n".join(
+            f"{msg['role']}: {msg['content']}" for msg in session.conversation
+        )
+        summary = await llm_service.generate(
+            messages=[{"role": "user", "content": f"Summarize this radio show call in 1-2 sentences:\n{transcript_text}"}],
+            system_prompt="You summarize radio show conversations concisely. Focus on what the caller talked about and any emotional moments.",
+        )
+
+    # Store in call history
+    session.call_history.append(CallRecord(
+        caller_type="real",
+        caller_name=caller_name,
+        summary=summary,
+        transcript=list(session.conversation),
+    ))
+
+    # Disconnect the caller
+    caller_service.hangup(caller_id)
+    await caller_service.disconnect_caller(caller_id)
+
+    # Stop host streaming if no more active callers
+    if len(caller_service.active_calls) == 0:
+        audio_service.stop_host_stream()
+
+    session.active_real_caller = None
+
+    # Play hangup sound
+    hangup_sound = settings.sounds_dir / "hangup.wav"
+    if hangup_sound.exists():
+        audio_service.play_sfx(str(hangup_sound))
+
+    # Auto follow-up?
+    auto_followup_triggered = False
+    if session.auto_followup:
+        auto_followup_triggered = True
+        asyncio.create_task(_auto_followup(summary))
+
+    return {
+        "status": "disconnected",
+        "caller": caller_name,
+        "summary": summary,
+        "auto_followup": auto_followup_triggered,
+    }
+
+
+async def _auto_followup(last_call_summary: str):
+    """Automatically pick an AI caller and connect them as follow-up"""
+    await asyncio.sleep(7)  # Brief pause before follow-up
+
+    # Ask LLM to pick best AI caller for follow-up
+    caller_list = ", ".join(
+        f'{k}: {v["name"]} ({v["gender"]}, {v["age_range"][0]}-{v["age_range"][1]})'
+        for k, v in CALLER_BASES.items()
+    )
+    pick = await llm_service.generate(
+        messages=[{"role": "user", "content": f'A caller just talked about: "{last_call_summary}". Which AI caller should follow up? Available: {caller_list}. Reply with just the key number.'}],
+        system_prompt="Pick the most interesting AI caller to follow up on this topic. Just reply with the number key.",
+    )
+
+    # Extract key from response
+    match = re.search(r'\d+', pick)
+    if match:
+        caller_key = match.group()
+        if caller_key in CALLER_BASES:
+            session.start_call(caller_key)
+            print(f"[Auto Follow-Up] {CALLER_BASES[caller_key]['name']} is calling in about: {last_call_summary[:50]}...")
+
+
+@app.post("/api/followup/generate")
+async def generate_followup():
+    """Generate an AI follow-up caller based on recent show history"""
+    if not session.call_history:
+        raise HTTPException(400, "No call history to follow up on")
+
+    last_record = session.call_history[-1]
+    await _auto_followup(last_record.summary)
+
+    return {
+        "status": "followup_triggered",
+        "based_on": last_record.caller_name,
+    }
+
+
+@app.post("/api/session/ai-mode")
+async def set_ai_mode(data: dict):
+    """Set AI respond mode (manual or auto)"""
+    mode = data.get("mode", "manual")
+    session.ai_respond_mode = mode
+    print(f"[Session] AI respond mode: {mode}")
+    return {"mode": mode}
+
+
+@app.post("/api/session/auto-followup")
+async def set_auto_followup(data: dict):
+    """Toggle auto follow-up"""
+    session.auto_followup = data.get("enabled", False)
+    print(f"[Session] Auto follow-up: {session.auto_followup}")
+    return {"enabled": session.auto_followup}
+
+
 # --- Server Control Endpoints ---
 
 import subprocess

backend/services/audio.py

@@ -48,6 +48,10 @@ class AudioService:
         self._caller_stop_event = threading.Event()
         self._caller_thread: Optional[threading.Thread] = None
 
+        # Host mic streaming state
+        self._host_stream: Optional[sd.InputStream] = None
+        self._host_send_callback: Optional[Callable] = None
+
         # Sample rates
         self.input_sample_rate = 16000  # For Whisper
         self.output_sample_rate = 24000  # For TTS
@@ -325,7 +329,7 @@ class AudioService:
             device_sr = int(device_info['default_samplerate'])
             channel_idx = min(channel, num_channels) - 1
 
-            # Resample from Twilio's 8kHz to device sample rate
+            # Resample to device sample rate if needed
             if sample_rate != device_sr:
                 audio = librosa.resample(audio, orig_sr=sample_rate, target_sr=device_sr)
 
@@ -345,6 +349,55 @@ class AudioService:
         except Exception as e:
             print(f"Real caller audio routing error: {e}")
 
+    # --- Host Mic Streaming ---
+
+    def start_host_stream(self, send_callback: Callable):
+        """Start continuous host mic capture for streaming to real callers"""
+        if self.input_device is None:
+            print("[Audio] No input device configured for host streaming")
+            return
+
+        self._host_send_callback = send_callback
+
+        device_info = sd.query_devices(self.input_device)
+        max_channels = device_info['max_input_channels']
+        device_sr = int(device_info['default_samplerate'])
+        record_channel = min(self.input_channel, max_channels) - 1
+
+        import librosa
+
+        def callback(indata, frames, time_info, status):
+            if not self._host_send_callback:
+                return
+            # Extract the configured input channel
+            mono = indata[:, record_channel].copy()
+            # Resample to 16kHz if needed
+            if device_sr != 16000:
+                mono = librosa.resample(mono, orig_sr=device_sr, target_sr=16000)
+            # Convert float32 to int16 PCM
+            pcm = (mono * 32767).astype(np.int16).tobytes()
+            self._host_send_callback(pcm)
+
+        self._host_stream = sd.InputStream(
+            device=self.input_device,
+            channels=max_channels,
+            samplerate=device_sr,
+            dtype=np.float32,
+            blocksize=4096,
+            callback=callback,
+        )
+        self._host_stream.start()
+        print(f"[Audio] Host mic streaming started (device {self.input_device} ch {self.input_channel} @ {device_sr}Hz)")
+
+    def stop_host_stream(self):
+        """Stop host mic streaming"""
+        if self._host_stream:
+            self._host_stream.stop()
+            self._host_stream.close()
+            self._host_stream = None
+            self._host_send_callback = None
+            print("[Audio] Host mic streaming stopped")
+
     # --- Music Playback ---
 
     def load_music(self, file_path: str) -> bool:

backend/services/caller_service.py

@@ -0,0 +1,147 @@
+"""Browser caller queue and audio stream service"""
+
+import asyncio
+import time
+import threading
+from typing import Optional
+
+
+class CallerService:
+    """Manages browser caller queue, channel allocation, and WebSocket streams"""
+
+    FIRST_REAL_CHANNEL = 3
+
+    def __init__(self):
+        self._queue: list[dict] = []
+        self.active_calls: dict[str, dict] = {}
+        self._allocated_channels: set[int] = set()
+        self._caller_counter: int = 0
+        self._lock = threading.Lock()
+        self._websockets: dict[str, any] = {}  # caller_id -> WebSocket
+
+    def add_to_queue(self, caller_id: str, name: str):
+        with self._lock:
+            self._queue.append({
+                "caller_id": caller_id,
+                "name": name,
+                "queued_at": time.time(),
+            })
+        print(f"[Caller] {name} added to queue (ID: {caller_id})")
+
+    def remove_from_queue(self, caller_id: str):
+        with self._lock:
+            self._queue = [c for c in self._queue if c["caller_id"] != caller_id]
+        print(f"[Caller] {caller_id} removed from queue")
+
+    def get_queue(self) -> list[dict]:
+        now = time.time()
+        with self._lock:
+            return [
+                {
+                    "caller_id": c["caller_id"],
+                    "name": c["name"],
+                    "wait_time": int(now - c["queued_at"]),
+                }
+                for c in self._queue
+            ]
+
+    def allocate_channel(self) -> int:
+        with self._lock:
+            ch = self.FIRST_REAL_CHANNEL
+            while ch in self._allocated_channels:
+                ch += 1
+            self._allocated_channels.add(ch)
+            return ch
+
+    def release_channel(self, channel: int):
+        with self._lock:
+            self._allocated_channels.discard(channel)
+
+    def take_call(self, caller_id: str) -> dict:
+        caller = None
+        with self._lock:
+            for c in self._queue:
+                if c["caller_id"] == caller_id:
+                    caller = c
+                    break
+            if caller:
+                self._queue = [c for c in self._queue if c["caller_id"] != caller_id]
+
+        if not caller:
+            raise ValueError(f"Caller {caller_id} not in queue")
+
+        channel = self.allocate_channel()
+        self._caller_counter += 1
+        name = caller["name"]
+
+        call_info = {
+            "caller_id": caller_id,
+            "name": name,
+            "channel": channel,
+            "started_at": time.time(),
+        }
+        self.active_calls[caller_id] = call_info
+        print(f"[Caller] {name} taken on air — channel {channel}")
+        return call_info
+
+    def hangup(self, caller_id: str):
+        call_info = self.active_calls.pop(caller_id, None)
+        if call_info:
+            self.release_channel(call_info["channel"])
+            print(f"[Caller] {call_info['name']} hung up — channel {call_info['channel']} released")
+        self._websockets.pop(caller_id, None)
+
+    def reset(self):
+        with self._lock:
+            for call_info in self.active_calls.values():
+                self._allocated_channels.discard(call_info["channel"])
+            self._queue.clear()
+            self.active_calls.clear()
+            self._allocated_channels.clear()
+            self._caller_counter = 0
+            self._websockets.clear()
+        print("[Caller] Service reset")
+
+    def register_websocket(self, caller_id: str, websocket):
+        """Register a WebSocket for a caller"""
+        self._websockets[caller_id] = websocket
+
+    def unregister_websocket(self, caller_id: str):
+        """Unregister a WebSocket"""
+        self._websockets.pop(caller_id, None)
+
+    async def send_audio_to_caller(self, caller_id: str, pcm_data: bytes, sample_rate: int):
+        """Send audio to real caller via WebSocket binary frame"""
+        ws = self._websockets.get(caller_id)
+        if not ws:
+            return
+
+        try:
+            if sample_rate != 16000:
+                import numpy as np
+                import librosa
+                audio = np.frombuffer(pcm_data, dtype=np.int16).astype(np.float32) / 32768.0
+                audio = librosa.resample(audio, orig_sr=sample_rate, target_sr=16000)
+                pcm_data = (audio * 32767).astype(np.int16).tobytes()
+            await ws.send_bytes(pcm_data)
+        except Exception as e:
+            print(f"[Caller] Failed to send audio: {e}")
+
+    async def notify_caller(self, caller_id: str, message: dict):
+        """Send JSON control message to caller"""
+        ws = self._websockets.get(caller_id)
+        if ws:
+            import json
+            await ws.send_text(json.dumps(message))
+
+    async def disconnect_caller(self, caller_id: str):
+        """Disconnect a caller's WebSocket"""
+        ws = self._websockets.get(caller_id)
+        if ws:
+            try:
+                import json
+                await ws.send_text(json.dumps({"status": "disconnected"}))
+                await ws.close()
+            except Exception:
+                pass
+        self._websockets.pop(caller_id, None)

backend/services/twilio_service.py

@@ -1,148 +0,0 @@
-"""Twilio call queue and media stream service"""
-
-import asyncio
-import base64
-import audioop
-import time
-import threading
-from typing import Optional
-
-
-class TwilioService:
-    """Manages Twilio call queue, channel allocation, and media streams"""
-
-    FIRST_REAL_CHANNEL = 3
-
-    def __init__(self):
-        self._queue: list[dict] = []
-        self.active_calls: dict[str, dict] = {}
-        self._allocated_channels: set[int] = set()
-        self._caller_counter: int = 0
-        self._lock = threading.Lock()
-        self._websockets: dict[str, any] = {}  # call_sid -> WebSocket
-
-    def add_to_queue(self, call_sid: str, phone: str):
-        with self._lock:
-            self._queue.append({
-                "call_sid": call_sid,
-                "phone": phone,
-                "queued_at": time.time(),
-            })
-        print(f"[Twilio] Caller {phone} added to queue (SID: {call_sid})")
-
-    def remove_from_queue(self, call_sid: str):
-        with self._lock:
-            self._queue = [c for c in self._queue if c["call_sid"] != call_sid]
-        print(f"[Twilio] Caller {call_sid} removed from queue")
-
-    def get_queue(self) -> list[dict]:
-        now = time.time()
-        with self._lock:
-            return [
-                {
-                    "call_sid": c["call_sid"],
-                    "phone": c["phone"],
-                    "wait_time": int(now - c["queued_at"]),
-                }
-                for c in self._queue
-            ]
-
-    def allocate_channel(self) -> int:
-        with self._lock:
-            ch = self.FIRST_REAL_CHANNEL
-            while ch in self._allocated_channels:
-                ch += 1
-            self._allocated_channels.add(ch)
-            return ch
-
-    def release_channel(self, channel: int):
-        with self._lock:
-            self._allocated_channels.discard(channel)
-
-    def take_call(self, call_sid: str) -> dict:
-        caller = None
-        with self._lock:
-            for c in self._queue:
-                if c["call_sid"] == call_sid:
-                    caller = c
-                    break
-            if caller:
-                self._queue = [c for c in self._queue if c["call_sid"] != call_sid]
-
-        if not caller:
-            raise ValueError(f"Call {call_sid} not in queue")
-
-        channel = self.allocate_channel()
-        self._caller_counter += 1
-        name = f"Caller #{self._caller_counter}"
-
-        call_info = {
-            "call_sid": call_sid,
-            "phone": caller["phone"],
-            "channel": channel,
-            "name": name,
-            "started_at": time.time(),
-        }
-        self.active_calls[call_sid] = call_info
-        print(f"[Twilio] {name} ({caller['phone']}) taken on air — channel {channel}")
-        return call_info
-
-    def hangup(self, call_sid: str):
-        call_info = self.active_calls.pop(call_sid, None)
-        if call_info:
-            self.release_channel(call_info["channel"])
-            print(f"[Twilio] {call_info['name']} hung up — channel {call_info['channel']} released")
-        self._websockets.pop(call_sid, None)
-
-    def reset(self):
-        with self._lock:
-            for call_info in self.active_calls.values():
-                self._allocated_channels.discard(call_info["channel"])
-            self._queue.clear()
-            self.active_calls.clear()
-            self._allocated_channels.clear()
-            self._caller_counter = 0
-            self._websockets.clear()
-        print("[Twilio] Service reset")
-
-    def register_websocket(self, call_sid: str, websocket):
-        """Register a WebSocket for a call"""
-        self._websockets[call_sid] = websocket
-
-    def unregister_websocket(self, call_sid: str):
-        """Unregister a WebSocket"""
-        self._websockets.pop(call_sid, None)
-
-    async def send_audio_to_caller(self, call_sid: str, pcm_data: bytes, sample_rate: int):
-        """Send audio back to real caller via Twilio WebSocket"""
-        ws = self._websockets.get(call_sid)
-        if not ws:
-            return
-
-        call_info = self.active_calls.get(call_sid)
-        if not call_info or "stream_sid" not in call_info:
-            return
-
-        try:
-            # Resample to 8kHz if needed
-            if sample_rate != 8000:
-                import numpy as np
-                import librosa
-                audio = np.frombuffer(pcm_data, dtype=np.int16).astype(np.float32) / 32768.0
-                audio = librosa.resample(audio, orig_sr=sample_rate, target_sr=8000)
-                pcm_data = (audio * 32767).astype(np.int16).tobytes()
-
-            # Convert PCM to mulaw
-            mulaw_data = audioop.lin2ulaw(pcm_data, 2)
-
-            # Send as Twilio media message
-            import json
-            await ws.send_text(json.dumps({
-                "event": "media",
-                "streamSid": call_info["stream_sid"],
-                "media": {
-                    "payload": base64.b64encode(mulaw_data).decode("ascii"),
-                },
-            }))
-        except Exception as e:
-            print(f"[Twilio] Failed to send audio to caller: {e}")

frontend/call-in.html

@@ -0,0 +1,155 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>Call In - Luke at the Roost</title>
+    <style>
+        :root {
+            --bg: #1a1a2e;
+            --bg-light: #252547;
+            --accent: #e94560;
+            --text: #fff;
+            --text-muted: #888;
+            --radius: 8px;
+        }
+        * { box-sizing: border-box; margin: 0; padding: 0; }
+        body {
+            font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif;
+            background: var(--bg);
+            color: var(--text);
+            min-height: 100vh;
+            display: flex;
+            align-items: center;
+            justify-content: center;
+        }
+        .container {
+            max-width: 400px;
+            width: 100%;
+            padding: 30px;
+            text-align: center;
+        }
+        h1 {
+            font-size: 1.5em;
+            margin-bottom: 8px;
+        }
+        .subtitle {
+            color: var(--text-muted);
+            margin-bottom: 30px;
+            font-size: 0.9em;
+        }
+        .form-group {
+            margin-bottom: 20px;
+        }
+        .form-group input {
+            width: 100%;
+            padding: 12px 16px;
+            border: 1px solid #444;
+            border-radius: var(--radius);
+            background: var(--bg-light);
+            color: var(--text);
+            font-size: 1em;
+            outline: none;
+        }
+        .form-group input:focus {
+            border-color: var(--accent);
+        }
+        .call-btn {
+            width: 100%;
+            padding: 14px;
+            border: none;
+            border-radius: var(--radius);
+            background: var(--accent);
+            color: #fff;
+            font-size: 1.1em;
+            font-weight: 600;
+            cursor: pointer;
+            transition: opacity 0.2s;
+        }
+        .call-btn:hover { opacity: 0.9; }
+        .call-btn:disabled { opacity: 0.5; cursor: not-allowed; }
+        .hangup-btn {
+            width: 100%;
+            padding: 14px;
+            border: none;
+            border-radius: var(--radius);
+            background: #c0392b;
+            color: #fff;
+            font-size: 1.1em;
+            font-weight: 600;
+            cursor: pointer;
+            margin-top: 12px;
+            display: none;
+        }
+        .hangup-btn:hover { opacity: 0.9; }
+        .status {
+            margin-top: 20px;
+            padding: 16px;
+            background: var(--bg-light);
+            border-radius: var(--radius);
+            display: none;
+        }
+        .status.visible { display: block; }
+        .status-label {
+            font-size: 0.85em;
+            color: var(--text-muted);
+            text-transform: uppercase;
+            letter-spacing: 1px;
+            margin-bottom: 6px;
+        }
+        .status-text {
+            font-size: 1.1em;
+            font-weight: 500;
+        }
+        .on-air .status-text {
+            color: var(--accent);
+            font-weight: 700;
+            animation: pulse 1.5s ease-in-out infinite;
+        }
+        @keyframes pulse {
+            0%, 100% { opacity: 1; }
+            50% { opacity: 0.6; }
+        }
+        .mic-meter {
+            margin-top: 16px;
+            height: 6px;
+            background: #333;
+            border-radius: 3px;
+            overflow: hidden;
+            display: none;
+        }
+        .mic-meter.visible { display: block; }
+        .mic-meter-fill {
+            height: 100%;
+            background: #2ecc71;
+            width: 0%;
+            transition: width 0.1s;
+            border-radius: 3px;
+        }
+    </style>
+</head>
+<body>
+    <div class="container">
+        <h1>Luke at the Roost</h1>
+        <p class="subtitle">Call in to the show</p>
+
+        <div class="form-group">
+            <input type="text" id="caller-name" placeholder="Your name" maxlength="30" autocomplete="off">
+        </div>
+
+        <button id="call-btn" class="call-btn">Call In</button>
+        <button id="hangup-btn" class="hangup-btn">Hang Up</button>
+
+        <div id="status" class="status">
+            <div class="status-label">Status</div>
+            <div id="status-text" class="status-text">Connecting...</div>
+        </div>
+
+        <div id="mic-meter" class="mic-meter">
+            <div id="mic-meter-fill" class="mic-meter-fill"></div>
+        </div>
+    </div>
+
+    <script src="/js/call-in.js"></script>
+</body>
+</html>

frontend/css/style.css

@@ -541,3 +541,38 @@ section h2 {
 .server-log .log-line.chat {
     color: #f8f;
 }
+
+/* Call Queue */
+.queue-section { margin: 1rem 0; }
+.call-queue { border: 1px solid #333; border-radius: 4px; padding: 0.5rem; max-height: 150px; overflow-y: auto; }
+.queue-empty { color: #666; text-align: center; padding: 0.5rem; }
+.queue-item { display: flex; align-items: center; gap: 0.75rem; padding: 0.4rem 0.5rem; border-bottom: 1px solid #222; }
+.queue-item:last-child { border-bottom: none; }
+.queue-phone { font-family: monospace; color: #4fc3f7; }
+.queue-wait { color: #999; font-size: 0.85rem; flex: 1; }
+.queue-take-btn { background: #2e7d32; color: white; border: none; padding: 0.25rem 0.75rem; border-radius: 3px; cursor: pointer; }
+.queue-take-btn:hover { background: #388e3c; }
+.queue-drop-btn { background: #c62828; color: white; border: none; padding: 0.25rem 0.5rem; border-radius: 3px; cursor: pointer; }
+.queue-drop-btn:hover { background: #d32f2f; }
+
+/* Active Call Indicator */
+.active-call { border: 1px solid #444; border-radius: 4px; padding: 0.75rem; margin: 0.5rem 0; background: #1a1a2e; }
+.caller-info { display: flex; align-items: center; gap: 0.5rem; margin-bottom: 0.5rem; }
+.caller-info:last-of-type { margin-bottom: 0; }
+.caller-type { font-size: 0.7rem; font-weight: bold; padding: 0.15rem 0.4rem; border-radius: 3px; text-transform: uppercase; }
+.caller-type.real { background: #c62828; color: white; }
+.caller-type.ai { background: #1565c0; color: white; }
+.channel-badge { font-size: 0.75rem; color: #999; background: #222; padding: 0.1rem 0.4rem; border-radius: 3px; }
+.call-duration { font-family: monospace; color: #4fc3f7; }
+.ai-controls { display: flex; align-items: center; gap: 0.5rem; margin-left: auto; }
+.mode-toggle { display: flex; border: 1px solid #444; border-radius: 3px; overflow: hidden; }
+.mode-btn { background: #222; color: #999; border: none; padding: 0.2rem 0.5rem; font-size: 0.75rem; cursor: pointer; }
+.mode-btn.active { background: #1565c0; color: white; }
+.respond-btn { background: #2e7d32; color: white; border: none; padding: 0.25rem 0.75rem; border-radius: 3px; font-size: 0.8rem; cursor: pointer; }
+.hangup-btn.small { font-size: 0.75rem; padding: 0.2rem 0.5rem; }
+.auto-followup-label { display: flex; align-items: center; gap: 0.4rem; font-size: 0.8rem; color: #999; margin-top: 0.5rem; }
+
+/* Three-Party Chat */
+.message.real-caller { border-left: 3px solid #c62828; padding-left: 0.5rem; }
+.message.ai-caller { border-left: 3px solid #1565c0; padding-left: 0.5rem; }
+.message.host { border-left: 3px solid #2e7d32; padding-left: 0.5rem; }

frontend/index.html

@@ -21,11 +21,44 @@
             <section class="callers-section">
                 <h2>Callers <span id="session-id" class="session-id"></span></h2>
                 <div id="callers" class="caller-grid"></div>
+                <!-- Active Call Indicator -->
+                <div id="active-call" class="active-call hidden">
+                    <div id="real-caller-info" class="caller-info hidden">
+                        <span class="caller-type real">LIVE</span>
+                        <span id="real-caller-name"></span>
+                        <span id="real-caller-channel" class="channel-badge"></span>
+                        <span id="real-caller-duration" class="call-duration"></span>
+                        <button id="hangup-real-btn" class="hangup-btn small">Hang Up</button>
+                    </div>
+                    <div id="ai-caller-info" class="caller-info hidden">
+                        <span class="caller-type ai">AI</span>
+                        <span id="ai-caller-name"></span>
+                        <div class="ai-controls">
+                            <div class="mode-toggle">
+                                <button id="mode-manual" class="mode-btn active">Manual</button>
+                                <button id="mode-auto" class="mode-btn">Auto</button>
+                            </div>
+                            <button id="ai-respond-btn" class="respond-btn">Let them respond</button>
+                        </div>
+                        <button id="hangup-ai-btn" class="hangup-btn small">Hang Up</button>
+                    </div>
+                    <label class="auto-followup-label">
+                        <input type="checkbox" id="auto-followup"> Auto Follow-Up
+                    </label>
+                </div>
                 <div id="call-status" class="call-status">No active call</div>
                 <div id="caller-background" class="caller-background hidden"></div>
                 <button id="hangup-btn" class="hangup-btn" disabled>Hang Up</button>
             </section>
 
+            <!-- Call Queue -->
+            <section class="queue-section">
+                <h2>Incoming Calls <a href="/call-in" target="_blank" style="font-size:0.6em;font-weight:normal;color:var(--accent);">Call-in page</a></h2>
+                <div id="call-queue" class="call-queue">
+                    <div class="queue-empty">No callers waiting</div>
+                </div>
+            </section>
+
             <!-- Chat -->
             <section class="chat-section">
                 <div id="chat" class="chat-log"></div>
@@ -173,6 +206,6 @@
         </div>
     </div>
 
-    <script src="/js/app.js?v=8"></script>
+    <script src="/js/app.js?v=9"></script>
 </body>
 </html>

frontend/js/app.js

@@ -52,6 +52,9 @@ function initEventListeners() {
     // Start log polling
     startLogPolling();
 
+    // Start queue polling
+    startQueuePolling();
+
     // Talk button - now triggers server-side recording
     const talkBtn = document.getElementById('talk-btn');
     if (talkBtn) {
@@ -97,6 +100,45 @@ function initEventListeners() {
         phoneFilter = e.target.checked;
     });
     document.getElementById('refresh-ollama')?.addEventListener('click', refreshOllamaModels);
+
+    // Real caller hangup
+    document.getElementById('hangup-real-btn')?.addEventListener('click', async () => {
+        await fetch('/api/hangup/real', { method: 'POST' });
+        hideRealCaller();
+        log('Real caller disconnected');
+    });
+
+    // AI respond mode toggle
+    document.getElementById('mode-manual')?.addEventListener('click', () => {
+        document.getElementById('mode-manual')?.classList.add('active');
+        document.getElementById('mode-auto')?.classList.remove('active');
+        document.getElementById('ai-respond-btn')?.classList.remove('hidden');
+        fetch('/api/session/ai-mode', {
+            method: 'POST',
+            headers: { 'Content-Type': 'application/json' },
+            body: JSON.stringify({ mode: 'manual' }),
+        });
+    });
+
+    document.getElementById('mode-auto')?.addEventListener('click', () => {
+        document.getElementById('mode-auto')?.classList.add('active');
+        document.getElementById('mode-manual')?.classList.remove('active');
+        document.getElementById('ai-respond-btn')?.classList.add('hidden');
+        fetch('/api/session/ai-mode', {
+            method: 'POST',
+            headers: { 'Content-Type': 'application/json' },
+            body: JSON.stringify({ mode: 'auto' }),
+        });
+    });
+
+    // Auto follow-up toggle
+    document.getElementById('auto-followup')?.addEventListener('change', (e) => {
+        fetch('/api/session/auto-followup', {
+            method: 'POST',
+            headers: { 'Content-Type': 'application/json' },
+            body: JSON.stringify({ enabled: e.target.checked }),
+        });
+    });
 }
 
 
@@ -273,9 +315,24 @@ async function startCall(key, name) {
 
     currentCaller = { key, name };
 
-    document.getElementById('call-status').textContent = `On call: ${name}`;
+    // Check if real caller is active (three-way scenario)
+    const realCallerActive = document.getElementById('real-caller-info') &&
+        !document.getElementById('real-caller-info').classList.contains('hidden');
+
+    if (realCallerActive) {
+        document.getElementById('call-status').textContent = `Three-way: ${name} (AI) + Real Caller`;
+    } else {
+        document.getElementById('call-status').textContent = `On call: ${name}`;
+    }
+
     document.getElementById('hangup-btn').disabled = false;
 
+    // Show AI caller in active call indicator
+    const aiInfo = document.getElementById('ai-caller-info');
+    const aiName = document.getElementById('ai-caller-name');
+    if (aiInfo) aiInfo.classList.remove('hidden');
+    if (aiName) aiName.textContent = name;
+
     // Show caller background
     const bgEl = document.getElementById('caller-background');
     if (bgEl && data.background) {
@@ -287,8 +344,10 @@ async function startCall(key, name) {
         btn.classList.toggle('active', btn.dataset.key === key);
     });
 
-    log(`Connected to ${name}`);
-    clearChat();
+    log(`Connected to ${name}` + (realCallerActive ? ' (three-way)' : ''));
+    if (!realCallerActive) clearChat();
+
+    updateActiveCallIndicator();
 }
 
 
@@ -314,7 +373,6 @@ async function newSession() {
 async function hangup() {
     if (!currentCaller) return;
 
-    // Stop any playing TTS
     await fetch('/api/tts/stop', { method: 'POST' });
     await fetch('/api/hangup', { method: 'POST' });
 
@@ -331,6 +389,10 @@ async function hangup() {
     // Hide caller background
     const bgEl = document.getElementById('caller-background');
     if (bgEl) bgEl.classList.add('hidden');
+
+    // Hide AI caller indicator
+    document.getElementById('ai-caller-info')?.classList.add('hidden');
+    updateActiveCallIndicator();
 }
 
 
@@ -647,7 +709,19 @@ function addMessage(sender, text) {
         return;
     }
     const div = document.createElement('div');
-    div.className = `message ${sender === 'You' ? 'host' : 'caller'}`;
+
+    let className = 'message';
+    if (sender === 'You') {
+        className += ' host';
+    } else if (sender === 'System') {
+        className += ' system';
+    } else if (sender.includes('(caller)') || sender.includes('Caller #')) {
+        className += ' real-caller';
+    } else {
+        className += ' ai-caller';
+    }
+
+    div.className = className;
     div.innerHTML = `<strong>${sender}:</strong> ${text}`;
     chat.appendChild(div);
     chat.scrollTop = chat.scrollHeight;
@@ -769,6 +843,121 @@ async function restartServer() {
 }
 
 
+// --- Call Queue ---
+let queuePollInterval = null;
+
+function startQueuePolling() {
+    queuePollInterval = setInterval(fetchQueue, 3000);
+    fetchQueue();
+}
+
+async function fetchQueue() {
+    try {
+        const res = await fetch('/api/queue');
+        const data = await res.json();
+        renderQueue(data.queue);
+    } catch (err) {}
+}
+
+function renderQueue(queue) {
+    const el = document.getElementById('call-queue');
+    if (!el) return;
+
+    if (queue.length === 0) {
+        el.innerHTML = '<div class="queue-empty">No callers waiting</div>';
+        return;
+    }
+
+    el.innerHTML = queue.map(caller => {
+        const mins = Math.floor(caller.wait_time / 60);
+        const secs = caller.wait_time % 60;
+        const waitStr = mins > 0 ? `${mins}m ${secs}s` : `${secs}s`;
+        return `
+            <div class="queue-item">
+                <span class="queue-name">${caller.name}</span>
+                <span class="queue-wait">waiting ${waitStr}</span>
+                <button class="queue-take-btn" onclick="takeCall('${caller.caller_id}')">Take Call</button>
+                <button class="queue-drop-btn" onclick="dropCall('${caller.caller_id}')">Drop</button>
+            </div>
+        `;
+    }).join('');
+}
+
+async function takeCall(callerId) {
+    try {
+        const res = await fetch(`/api/queue/take/${callerId}`, { method: 'POST' });
+        const data = await res.json();
+        if (data.status === 'on_air') {
+            showRealCaller(data.caller);
+            log(`${data.caller.name} is on air — Channel ${data.caller.channel}`);
+        }
+    } catch (err) {
+        log('Failed to take call: ' + err.message);
+    }
+}
+
+async function dropCall(callerId) {
+    try {
+        await fetch(`/api/queue/drop/${callerId}`, { method: 'POST' });
+        fetchQueue();
+    } catch (err) {
+        log('Failed to drop call: ' + err.message);
+    }
+}
+
+
+// --- Active Call Indicator ---
+let realCallerTimer = null;
+let realCallerStartTime = null;
+
+function updateActiveCallIndicator() {
+    const container = document.getElementById('active-call');
+    const realInfo = document.getElementById('real-caller-info');
+    const aiInfo = document.getElementById('ai-caller-info');
+    const statusEl = document.getElementById('call-status');
+
+    const hasReal = realInfo && !realInfo.classList.contains('hidden');
+    const hasAi = aiInfo && !aiInfo.classList.contains('hidden');
+
+    if (hasReal || hasAi) {
+        container?.classList.remove('hidden');
+        statusEl?.classList.add('hidden');
+    } else {
+        container?.classList.add('hidden');
+        statusEl?.classList.remove('hidden');
+        if (statusEl) statusEl.textContent = 'No active call';
+    }
+}
+
+function showRealCaller(callerInfo) {
+    const nameEl = document.getElementById('real-caller-name');
+    const chEl = document.getElementById('real-caller-channel');
+    if (nameEl) nameEl.textContent = callerInfo.name;
+    if (chEl) chEl.textContent = `Ch ${callerInfo.channel}`;
+
+    document.getElementById('real-caller-info')?.classList.remove('hidden');
+    realCallerStartTime = Date.now();
+
+    if (realCallerTimer) clearInterval(realCallerTimer);
+    realCallerTimer = setInterval(() => {
+        const elapsed = Math.floor((Date.now() - realCallerStartTime) / 1000);
+        const mins = Math.floor(elapsed / 60);
+        const secs = elapsed % 60;
+        const durEl = document.getElementById('real-caller-duration');
+        if (durEl) durEl.textContent = `${mins}:${secs.toString().padStart(2, '0')}`;
+    }, 1000);
+
+    updateActiveCallIndicator();
+}
+
+function hideRealCaller() {
+    document.getElementById('real-caller-info')?.classList.add('hidden');
+    if (realCallerTimer) clearInterval(realCallerTimer);
+    realCallerTimer = null;
+    updateActiveCallIndicator();
+}
+
+
 async function stopServer() {
     if (!confirm('Stop the server? You will need to restart it manually.')) return;
 

frontend/js/call-in.js

@@ -0,0 +1,232 @@
+/**
+ * Call-In Page — Browser WebSocket audio streaming
+ * Captures mic via AudioWorklet, sends Int16 PCM 16kHz mono over WebSocket.
+ * Receives Int16 PCM 16kHz mono back for playback.
+ */
+
+let ws = null;
+let audioCtx = null;
+let micStream = null;
+let workletNode = null;
+let nextPlayTime = 0;
+let callerId = null;
+
+const callBtn = document.getElementById('call-btn');
+const hangupBtn = document.getElementById('hangup-btn');
+const statusEl = document.getElementById('status');
+const statusText = document.getElementById('status-text');
+const nameInput = document.getElementById('caller-name');
+const micMeter = document.getElementById('mic-meter');
+const micMeterFill = document.getElementById('mic-meter-fill');
+
+callBtn.addEventListener('click', startCall);
+hangupBtn.addEventListener('click', hangUp);
+nameInput.addEventListener('keydown', e => {
+    if (e.key === 'Enter') startCall();
+});
+
+async function startCall() {
+    const name = nameInput.value.trim() || 'Anonymous';
+    callBtn.disabled = true;
+    setStatus('Connecting...', false);
+
+    try {
+        // Get mic access
+        micStream = await navigator.mediaDevices.getUserMedia({
+            audio: { echoCancellation: true, noiseSuppression: true, sampleRate: 16000 }
+        });
+
+        // Set up AudioContext
+        audioCtx = new AudioContext({ sampleRate: 48000 });
+
+        // Register worklet processor inline via blob
+        const processorCode = `
+class CallerProcessor extends AudioWorkletProcessor {
+    constructor() {
+        super();
+        this.buffer = [];
+        this.targetSamples = 4096; // ~256ms at 16kHz
+    }
+    process(inputs) {
+        const input = inputs[0][0];
+        if (!input) return true;
+
+        // Downsample from sampleRate to 16000
+        const ratio = sampleRate / 16000;
+        for (let i = 0; i < input.length; i += ratio) {
+            const idx = Math.floor(i);
+            if (idx < input.length) {
+                this.buffer.push(input[idx]);
+            }
+        }
+
+        if (this.buffer.length >= this.targetSamples) {
+            const chunk = this.buffer.splice(0, this.targetSamples);
+            const int16 = new Int16Array(chunk.length);
+            for (let i = 0; i < chunk.length; i++) {
+                const s = Math.max(-1, Math.min(1, chunk[i]));
+                int16[i] = s < 0 ? s * 32768 : s * 32767;
+            }
+            this.port.postMessage(int16.buffer, [int16.buffer]);
+        }
+        return true;
+    }
+}
+registerProcessor('caller-processor', CallerProcessor);
+`;
+        const blob = new Blob([processorCode], { type: 'application/javascript' });
+        const blobUrl = URL.createObjectURL(blob);
+        await audioCtx.audioWorklet.addModule(blobUrl);
+        URL.revokeObjectURL(blobUrl);
+
+        // Connect mic to worklet
+        const source = audioCtx.createMediaStreamSource(micStream);
+        workletNode = new AudioWorkletNode(audioCtx, 'caller-processor');
+
+        // Connect WebSocket
+        const proto = location.protocol === 'https:' ? 'wss:' : 'ws:';
+        ws = new WebSocket(`${proto}//${location.host}/api/caller/stream`);
+        ws.binaryType = 'arraybuffer';
+
+        ws.onopen = () => {
+            ws.send(JSON.stringify({ type: 'join', name }));
+        };
+
+        ws.onmessage = (event) => {
+            if (typeof event.data === 'string') {
+                handleControlMessage(JSON.parse(event.data));
+            } else {
+                handleAudioData(event.data);
+            }
+        };
+
+        ws.onclose = () => {
+            setStatus('Disconnected', false);
+            cleanup();
+        };
+
+        ws.onerror = () => {
+            setStatus('Connection error', false);
+            cleanup();
+        };
+
+        // Forward mic audio to WebSocket
+        workletNode.port.onmessage = (e) => {
+            if (ws && ws.readyState === WebSocket.OPEN) {
+                ws.send(e.data);
+            }
+        };
+
+        source.connect(workletNode);
+        // Don't connect worklet to destination — we don't want to hear our own mic
+
+        // Show mic meter
+        const analyser = audioCtx.createAnalyser();
+        analyser.fftSize = 256;
+        source.connect(analyser);
+        startMicMeter(analyser);
+
+        // UI
+        nameInput.disabled = true;
+        hangupBtn.style.display = 'block';
+
+    } catch (err) {
+        console.error('Call error:', err);
+        setStatus('Failed: ' + err.message, false);
+        callBtn.disabled = false;
+        cleanup();
+    }
+}
+
+function handleControlMessage(msg) {
+    if (msg.status === 'queued') {
+        callerId = msg.caller_id;
+        setStatus(`Waiting in queue (position ${msg.position})...`, false);
+    } else if (msg.status === 'on_air') {
+        setStatus('ON AIR', true);
+        nextPlayTime = audioCtx.currentTime;
+    } else if (msg.status === 'disconnected') {
+        setStatus('Disconnected', false);
+        cleanup();
+    }
+}
+
+function handleAudioData(buffer) {
+    if (!audioCtx) return;
+
+    const int16 = new Int16Array(buffer);
+    const float32 = new Float32Array(int16.length);
+    for (let i = 0; i < int16.length; i++) {
+        float32[i] = int16[i] / 32768;
+    }
+
+    const audioBuf = audioCtx.createBuffer(1, float32.length, 16000);
+    audioBuf.copyToChannel(float32, 0);
+
+    const source = audioCtx.createBufferSource();
+    source.buffer = audioBuf;
+    source.connect(audioCtx.destination);
+
+    const now = audioCtx.currentTime;
+    if (nextPlayTime < now) {
+        nextPlayTime = now;
+    }
+    source.start(nextPlayTime);
+    nextPlayTime += audioBuf.duration;
+}
+
+function hangUp() {
+    if (ws && ws.readyState === WebSocket.OPEN) {
+        ws.close();
+    }
+    setStatus('Disconnected', false);
+    cleanup();
+}
+
+function cleanup() {
+    if (workletNode) {
+        workletNode.disconnect();
+        workletNode = null;
+    }
+    if (micStream) {
+        micStream.getTracks().forEach(t => t.stop());
+        micStream = null;
+    }
+    if (audioCtx) {
+        audioCtx.close().catch(() => {});
+        audioCtx = null;
+    }
+    ws = null;
+    callerId = null;
+    callBtn.disabled = false;
+    nameInput.disabled = false;
+    hangupBtn.style.display = 'none';
+    micMeter.classList.remove('visible');
+}
+
+function setStatus(text, isOnAir) {
+    statusEl.classList.add('visible');
+    statusText.textContent = text;
+    if (isOnAir) {
+        statusEl.classList.add('on-air');
+    } else {
+        statusEl.classList.remove('on-air');
+    }
+}
+
+function startMicMeter(analyser) {
+    micMeter.classList.add('visible');
+    const data = new Uint8Array(analyser.frequencyBinCount);
+
+    function update() {
+        if (!analyser || !audioCtx) return;
+        analyser.getByteFrequencyData(data);
+        let sum = 0;
+        for (let i = 0; i < data.length; i++) sum += data[i];
+        const avg = sum / data.length;
+        const pct = Math.min(100, (avg / 128) * 100);
+        micMeterFill.style.width = pct + '%';
+        requestAnimationFrame(update);
+    }
+    update();
+}

tests/test_caller_service.py

@@ -0,0 +1,180 @@
+import sys
+sys.path.insert(0, "/Users/lukemacneil/ai-podcast")
+
+from backend.services.caller_service import CallerService
+
+
+def test_queue_starts_empty():
+    svc = CallerService()
+    assert svc.get_queue() == []
+
+
+def test_add_caller_to_queue():
+    svc = CallerService()
+    svc.add_to_queue("abc123", "Dave")
+    q = svc.get_queue()
+    assert len(q) == 1
+    assert q[0]["caller_id"] == "abc123"
+    assert q[0]["name"] == "Dave"
+    assert "wait_time" in q[0]
+
+
+def test_remove_caller_from_queue():
+    svc = CallerService()
+    svc.add_to_queue("abc123", "Dave")
+    svc.remove_from_queue("abc123")
+    assert svc.get_queue() == []
+
+
+def test_allocate_channel():
+    svc = CallerService()
+    ch1 = svc.allocate_channel()
+    ch2 = svc.allocate_channel()
+    assert ch1 == 3
+    assert ch2 == 4
+    svc.release_channel(ch1)
+    ch3 = svc.allocate_channel()
+    assert ch3 == 3
+
+
+def test_take_call():
+    svc = CallerService()
+    svc.add_to_queue("abc123", "Dave")
+    result = svc.take_call("abc123")
+    assert result["caller_id"] == "abc123"
+    assert result["channel"] >= 3
+    assert svc.get_queue() == []
+    assert svc.active_calls["abc123"]["channel"] == result["channel"]
+
+
+def test_hangup_real_caller():
+    svc = CallerService()
+    svc.add_to_queue("abc123", "Dave")
+    svc.take_call("abc123")
+    ch = svc.active_calls["abc123"]["channel"]
+    svc.hangup("abc123")
+    assert "abc123" not in svc.active_calls
+    assert ch not in svc._allocated_channels
+
+
+def test_caller_counter_increments():
+    svc = CallerService()
+    svc.add_to_queue("id1", "Dave")
+    svc.add_to_queue("id2", "Sarah")
+    r1 = svc.take_call("id1")
+    r2 = svc.take_call("id2")
+    assert r1["name"] == "Dave"
+    assert r2["name"] == "Sarah"
+
+
+def test_register_and_unregister_websocket():
+    svc = CallerService()
+    fake_ws = object()
+    svc.register_websocket("abc123", fake_ws)
+    assert svc._websockets["abc123"] is fake_ws
+    svc.unregister_websocket("abc123")
+    assert "abc123" not in svc._websockets
+
+
+def test_hangup_clears_websocket():
+    svc = CallerService()
+    svc.add_to_queue("abc123", "Dave")
+    svc.take_call("abc123")
+    svc.register_websocket("abc123", object())
+    svc.hangup("abc123")
+    assert "abc123" not in svc._websockets
+
+
+def test_reset_clears_websockets():
+    svc = CallerService()
+    svc.register_websocket("id1", object())
+    svc.register_websocket("id2", object())
+    svc.reset()
+    assert svc._websockets == {}
+
+
+def test_send_audio_no_websocket():
+    """send_audio_to_caller returns silently when no WS registered"""
+    import asyncio
+    svc = CallerService()
+    asyncio.get_event_loop().run_until_complete(
+        svc.send_audio_to_caller("NONE", b"\x00" * 100, 16000)
+    )
+
+
+def test_notify_caller():
+    """notify_caller sends JSON text to WebSocket"""
+    import asyncio
+
+    class FakeWS:
+        def __init__(self):
+            self.sent = []
+        async def send_text(self, data):
+            self.sent.append(data)
+
+    svc = CallerService()
+    ws = FakeWS()
+    svc.register_websocket("abc123", ws)
+    asyncio.get_event_loop().run_until_complete(
+        svc.notify_caller("abc123", {"status": "on_air", "channel": 3})
+    )
+    assert len(ws.sent) == 1
+    import json
+    msg = json.loads(ws.sent[0])
+    assert msg["status"] == "on_air"
+    assert msg["channel"] == 3
+
+
+def test_disconnect_caller():
+    """disconnect_caller sends disconnected message and removes WS"""
+    import asyncio
+
+    class FakeWS:
+        def __init__(self):
+            self.sent = []
+            self.closed = False
+        async def send_text(self, data):
+            self.sent.append(data)
+        async def close(self):
+            self.closed = True
+
+    svc = CallerService()
+    ws = FakeWS()
+    svc.register_websocket("abc123", ws)
+    asyncio.get_event_loop().run_until_complete(
+        svc.disconnect_caller("abc123")
+    )
+    assert ws.closed
+    assert "abc123" not in svc._websockets
+    import json
+    msg = json.loads(ws.sent[0])
+    assert msg["status"] == "disconnected"
+
+
+def test_send_audio_binary():
+    """send_audio_to_caller sends raw PCM bytes (not mulaw/JSON)"""
+    import asyncio
+
+    class FakeWS:
+        def __init__(self):
+            self.sent_bytes = []
+        async def send_bytes(self, data):
+            self.sent_bytes.append(data)
+
+    svc = CallerService()
+    ws = FakeWS()
+    svc.register_websocket("abc123", ws)
+    pcm = b"\x00\x01" * 100
+    asyncio.get_event_loop().run_until_complete(
+        svc.send_audio_to_caller("abc123", pcm, 16000)
+    )
+    assert len(ws.sent_bytes) == 1
+    assert ws.sent_bytes[0] == pcm
+
+
+def test_take_call_preserves_caller_name():
+    """take_call uses the name from the queue, not a generic counter name"""
+    svc = CallerService()
+    svc.add_to_queue("abc123", "Dave from Chicago")
+    result = svc.take_call("abc123")
+    assert result["name"] == "Dave from Chicago"

tests/test_followup.py

@@ -0,0 +1,39 @@
+import sys
+sys.path.insert(0, "/Users/lukemacneil/ai-podcast")
+
+from backend.main import Session, CallRecord, get_caller_prompt
+
+
+def test_caller_prompt_includes_show_history():
+    s = Session()
+    s.call_history.append(CallRecord(
+        caller_type="real", caller_name="Dave",
+        summary="Called about his wife leaving after 12 years",
+        transcript=[],
+    ))
+
+    s.start_call("1")  # Tony
+    caller = s.caller
+    show_history = s.get_show_history()
+    prompt = get_caller_prompt(caller, "", show_history)
+    assert "Dave" in prompt
+    assert "wife leaving" in prompt
+    assert "EARLIER IN THE SHOW" in prompt
+
+
+def test_caller_prompt_without_history():
+    s = Session()
+    s.start_call("1")
+    caller = s.caller
+    prompt = get_caller_prompt(caller, "")
+    assert "EARLIER IN THE SHOW" not in prompt
+    assert caller["name"] in prompt
+
+
+def test_caller_prompt_backward_compatible():
+    """Verify get_caller_prompt works with just 2 args (no show_history)"""
+    s = Session()
+    s.start_call("1")
+    caller = s.caller
+    prompt = get_caller_prompt(caller, "Host: hello")
+    assert "hello" in prompt

tests/test_session.py

@@ -30,8 +30,8 @@ def test_session_active_real_caller():
     s = Session()
     assert s.active_real_caller is None
     s.active_real_caller = {
-        "call_sid": "CA123", "phone": "+15125550142",
-        "channel": 3, "name": "Caller #1",
+        "caller_id": "abc123",
+        "channel": 3, "name": "Dave",
     }
     assert s.active_real_caller["channel"] == 3
 
@@ -70,7 +70,7 @@ def test_session_reset_clears_history():
         caller_type="real", caller_name="Dave",
         summary="test", transcript=[],
     ))
-    s.active_real_caller = {"call_sid": "CA123"}
+    s.active_real_caller = {"caller_id": "abc123"}
     s.ai_respond_mode = "auto"
     s.reset()
     assert s.call_history == []

tests/test_twilio_service.py

@@ -1,103 +0,0 @@
-import sys
-sys.path.insert(0, "/Users/lukemacneil/ai-podcast")
-
-from backend.services.twilio_service import TwilioService
-
-
-def test_queue_starts_empty():
-    svc = TwilioService()
-    assert svc.get_queue() == []
-
-
-def test_add_caller_to_queue():
-    svc = TwilioService()
-    svc.add_to_queue("CA123", "+15125550142")
-    q = svc.get_queue()
-    assert len(q) == 1
-    assert q[0]["call_sid"] == "CA123"
-    assert q[0]["phone"] == "+15125550142"
-    assert "wait_time" in q[0]
-
-
-def test_remove_caller_from_queue():
-    svc = TwilioService()
-    svc.add_to_queue("CA123", "+15125550142")
-    svc.remove_from_queue("CA123")
-    assert svc.get_queue() == []
-
-
-def test_allocate_channel():
-    svc = TwilioService()
-    ch1 = svc.allocate_channel()
-    ch2 = svc.allocate_channel()
-    assert ch1 == 3
-    assert ch2 == 4
-    svc.release_channel(ch1)
-    ch3 = svc.allocate_channel()
-    assert ch3 == 3
-
-
-def test_take_call():
-    svc = TwilioService()
-    svc.add_to_queue("CA123", "+15125550142")
-    result = svc.take_call("CA123")
-    assert result["call_sid"] == "CA123"
-    assert result["channel"] >= 3
-    assert svc.get_queue() == []
-    assert svc.active_calls["CA123"]["channel"] == result["channel"]
-
-
-def test_hangup_real_caller():
-    svc = TwilioService()
-    svc.add_to_queue("CA123", "+15125550142")
-    svc.take_call("CA123")
-    ch = svc.active_calls["CA123"]["channel"]
-    svc.hangup("CA123")
-    assert "CA123" not in svc.active_calls
-    assert ch not in svc._allocated_channels
-
-
-def test_caller_counter_increments():
-    svc = TwilioService()
-    svc.add_to_queue("CA1", "+15125550001")
-    svc.add_to_queue("CA2", "+15125550002")
-    r1 = svc.take_call("CA1")
-    r2 = svc.take_call("CA2")
-    assert r1["name"] == "Caller #1"
-    assert r2["name"] == "Caller #2"
-
-
-def test_register_and_unregister_websocket():
-    svc = TwilioService()
-    fake_ws = object()
-    svc.register_websocket("CA123", fake_ws)
-    assert svc._websockets["CA123"] is fake_ws
-    svc.unregister_websocket("CA123")
-    assert "CA123" not in svc._websockets
-
-
-def test_hangup_clears_websocket():
-    svc = TwilioService()
-    svc.add_to_queue("CA123", "+15125550142")
-    svc.take_call("CA123")
-    svc.register_websocket("CA123", object())
-    svc.hangup("CA123")
-    assert "CA123" not in svc._websockets
-
-
-def test_reset_clears_websockets():
-    svc = TwilioService()
-    svc.register_websocket("CA1", object())
-    svc.register_websocket("CA2", object())
-    svc.reset()
-    assert svc._websockets == {}
-
-
-def test_send_audio_no_websocket():
-    """send_audio_to_caller returns silently when no WS registered"""
-    import asyncio
-    svc = TwilioService()
-    # Should not raise
-    asyncio.get_event_loop().run_until_complete(
-        svc.send_audio_to_caller("CA_NONE", b"\x00" * 100, 8000)
-    )