Voice or chat not working when running on Apple with docker

[HubertReX]

Hi 👋

This project looks awesome, and I've spent several hours trying to run it on my mac-mini (M4 24 GB). I think I'm close to making it work, but still, I can't get the agent to respond. I see that my voice is detected (not using wake word), tried also writing - same result - no response. I'm following the README-APPLE.md instructions. I'm using code. I've pulled code from this repo ~7h ago - last commit 310043ca6e78a41084b10448e87645b9d012222f. I've got all containers running, with Ollama, mlx-audio running on host. I've event got the tailscale configuration to work so I can access the web on https and the mic is enabled in the browser.

mlx-audi seems to work:

> tail -f /tmp/caal-mlx-audio.log
Fetching 2 files: 100%|██████████| 2/2 [00:00<00:00, 28244.47it/s]
Fetching 4 files: 100%|██████████| 4/4 [00:00<00:00, 44384.17it/s]
INFO:     127.0.0.1:58264 - "POST /v1/models?model_name=mlx-community/whisper-large-v3-turbo HTTP/1.1" 200 OK
Fetching 9 files: 100%|██████████| 9/9 [00:00<00:00, 15307.68it/s]
Fetching 9 files: 100%|██████████| 9/9 [00:00<00:00, 9434.83it/s]
INFO:     127.0.0.1:58270 - "POST /v1/models?model_name=prince-canuma/Kokoro-82M HTTP/1.1" 200 OK
INFO:     127.0.0.1:58290 - "POST /v1/models?model_name=mlx-community/whisper-large-v3-turbo HTTP/1.1" 200 OK

the webhook is responding on health endpoint:

curl -X GET http://localhost:8889/health \
  -H "Content-Type: application/json"
{"status":"ok","active_sessions":[]}%

but complains on the annonce endpoint:

curl -X POST http://localhost:8889/announce \
  -H "Content-Type: application/json" \
  -d '{"message": "Package delivered at front door"}'
{"detail":"No active session in room: voice_assistant_room"}%

caal- logs:

HTTPS mode: enabling TURN/TLS for mac-mini.kamori-vector.ts.net
2026-01-04T13:33:35.195Z	INFO	livekit	routing/interfaces.go:180	using single-node routing
2026-01-04T13:33:35.198Z	INFO	livekit	service/turn.go:145	Starting TURN server	{"turn.relay_range_start": 30000, "turn.relay_range_end": 30100, "turn.portTLS": 5349, "turn.externalTLS": true}
2026-01-04T13:33:35.199Z	INFO	livekit	service/server.go:264	starting LiveKit server	{"portHttp": 7880, "nodeID": "ND_vVJbthxwL66H", "nodeIP": "100.107.186.2", "version": "1.9.10", "rtc.portTCP": 7881, "rtc.portICERange": [50000, 50100]}
2026-01-04T13:33:51.160Z	INFO	livekit.agents	service/agentservice.go:298	worker registered	{"namespace": "", "jobType": "JT_ROOM", "agentName": "", "workerID": "AW_TKMenpwESCKQ"}
2026-01-04T13:34:04.327Z	INFO	livekit	service/roommanager.go:413	starting RTC session	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "voice_assistant_user_3923", "pID": "PA_gMjTYm6w8eZX", "remote": false, "room": "voice_assistant_room", "nodeID": "ND_vVJbthxwL66H", "numParticipants": 0, "participantInit": {"Identity": "voice_assistant_user_3923", "Reconnect": false, "ReconnectReason": "RR_UNKNOWN", "AutoSubscribe": true, "Client": {"sdk": "JS", "version": "2.15.15", "protocol": 16, "os": "Mac OS X", "osVersion": "10.15.7", "deviceModel": "Mac", "browser": "Chrome", "browserVersion": "143.0.0"}, "Grants": {"Identity": "voice_assistant_user_3923", "Kind": "", "KindDetails": [], "Video": {"RoomCreate": false, "RoomList": false, "RoomRecord": false, "RoomAdmin": false, "RoomJoin": true, "Room": "voice_assistant_room", "CanPublish": true, "CanSubscribe": true, "CanPublishData": true, "CanPublishSources": [], "CanUpdateOwnMetadata": "not-set", "IngressAdmin": false, "Hidden": false, "Recorder": false, "Agent": false, "CanSubscribeMetrics": "not-set", "DestinationRoom": ""}, "SIP": {}, "Agent": {}, "Inference": {}, "Observability": {}, "RoomConfig": {"departureTimeout": 1}, "RoomPreset": ""}, "Region": "", "AdaptiveStream": false, "ID": "", "SubscriberAllowPause": "not-set", "DisableICELite": false, "CreateRoom": {"name": "voice_assistant_room", "departureTimeout": 1}, "AddTrackRequests": [], "PublisherOffer": {}, "SyncState": {}, "UseSinglePeerConnection": false}}
2026-01-04T13:34:04.328Z	INFO	livekit.transport	rtc/transport.go:1197	data tracks not enabled	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "voice_assistant_user_3923", "pID": "PA_gMjTYm6w8eZX", "remote": false, "transport": "SUBSCRIBER"}
2026-01-04T13:34:04.328Z	INFO	livekit	service/roommanager.go:1003	created TURN password	{"username": "YpVZ4cnNtlFVq10ZfKIo4zumqLsZGSNhSjppgKJoCC", "password": "fK8JBYr88ZWuKfrldpDmLQzJcdfEMbEaVetbMp6cfT6F"}
2026-01-04T13:34:04.435Z	INFO	livekit	rtc/room.go:1249	participant active	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "voice_assistant_user_3923", "pID": "PA_gMjTYm6w8eZX", "remote": false, "publisherCandidates": ["[local][selected:1][trickle] udp4 host 100.107.186.2:50028 (resolved: 100.107.186.2:50028)", "[local][trickle] tcp4 host 100.107.186.2:7881 (resolved: 100.107.186.2:7881)", "[remote][trickle] udp4 host 192.168.1.75:58178", "[remote][selected:1][trickle] udp4 host 100.78.104...:49383", "[remote][trickle] udp6 host fd7a:115c:a1e0::5201:6818:60610", "[remote][trickle] udp4 host 192.168.1.75:51850", "[remote][trickle] udp4 host 100.78.104...:59934", "[remote][trickle] udp6 host fd7a:115c:a1e0::5201:6818:58982", "[remote][trickle] udp4 srflx 83.29.5...:58178  related 192.168.1.75:58178"], "subscriberCandidates": ["[local][selected:1][trickle] udp4 host 100.107.186.2:50071 (resolved: 100.107.186.2:50071)", "[local][trickle] tcp4 host 100.107.186.2:7881 (resolved: 100.107.186.2:7881)", "[remote][trickle] udp4 host 192.168.1.75:51979", "[remote][selected:1][trickle] udp4 host 100.78.104...:55773", "[remote][trickle] udp6 host fd7a:115c:a1e0::5201:6818:64450", "[remote][trickle] udp4 srflx 83.29.5...:51979  related 192.168.1.75:51979"], "connectionType": "udp", "clientInfo": {"sdk": "JS", "version": "2.15.15", "protocol": 16, "os": "Mac OS X", "osVersion": "10.15.7", "deviceModel": "Mac", "browser": "Chrome", "browserVersion": "143.0.0"}, "connectTime": "108.391668ms"}
2026-01-04T13:34:04.446Z	INFO	livekit.pub	rtc/participant.go:2244	mediaTrack published	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "voice_assistant_user_3923", "pID": "PA_gMjTYm6w8eZX", "remote": false, "kind": "audio", "trackID": "TR_AMjJatwp6vnciS", "webrtcTrackID": "581d5cf4-0970-4dad-a54c-5cdd37fe4ced", "rid": "", "ssrc": 445952574, "rtxSsrc": 0, "mime": "audio/red", "trackInfo": {"sid": "TR_AMjJatwp6vnciS", "type": "AUDIO", "source": "MICROPHONE", "mimeType": "audio/red", "mid": "0", "codecs": [{"mimeType": "audio/red", "mid": "0", "cid": "581d5cf4-0970-4dad-a54c-5cdd37fe4ced", "videoLayerMode": "MODE_UNUSED"}], "encryption": "NONE", "stream": "camera", "version": {"unixMicro": 1767533644446400}, "audioFeatures": ["TF_AUTO_GAIN_CONTROL", "TF_ECHO_CANCELLATION", "TF_NOISE_SUPPRESSION"], "backupCodecPolicy": "SIMULCAST"}, "fromSdp": true, "isReceiverAdded": true, "sdpRids": ["", "", ""]}
2026-01-04T13:34:04.529Z	INFO	livekit.agents	service/agentservice.go:407	assigned job to worker	{"jobID": "AJ_cAurGAkEUra2", "namespace": "", "agentName": "", "jobType": "JT_ROOM", "room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "workerID": "AW_TKMenpwESCKQ"}
2026-01-04T13:34:04.555Z	INFO	livekit	service/roommanager.go:413	starting RTC session	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "agent-AJ_cAurGAkEUra2", "pID": "PA_ApqyKp5yiCUL", "remote": false, "room": "voice_assistant_room", "nodeID": "ND_vVJbthxwL66H", "numParticipants": 1, "participantInit": {"Identity": "agent-AJ_cAurGAkEUra2", "Reconnect": false, "ReconnectReason": "RR_UNKNOWN", "AutoSubscribe": true, "Client": {"sdk": "PYTHON", "version": "1.0.19", "protocol": 16}, "Grants": {"Identity": "agent-AJ_cAurGAkEUra2", "Kind": "agent", "KindDetails": [], "Video": {"RoomCreate": false, "RoomList": false, "RoomRecord": false, "RoomAdmin": false, "RoomJoin": true, "Room": "voice_assistant_room", "CanPublish": true, "CanSubscribe": true, "CanPublishData": true, "CanPublishSources": [], "CanUpdateOwnMetadata": true, "IngressAdmin": false, "Hidden": false, "Recorder": false, "Agent": true, "CanSubscribeMetrics": false, "DestinationRoom": ""}, "SIP": {}, "Agent": {}, "Inference": {}, "Observability": {}, "RoomConfig": {}, "RoomPreset": ""}, "Region": "", "AdaptiveStream": false, "ID": "", "SubscriberAllowPause": "not-set", "DisableICELite": false, "CreateRoom": {"name": "voice_assistant_room"}, "AddTrackRequests": [], "PublisherOffer": {}, "SyncState": {}, "UseSinglePeerConnection": false}}
��
04T13:34:04.556Z	INFO	livekit.transport	rtc/transport.go:1197	data tracks not enabled	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "agent-AJ_cAurGAkEUra2", "pID": "PA_ApqyKp5yiCUL", "remote": false, "transport": "SUBSCRIBER"}
2026-01-04T13:34:04.556Z	INFO	livekit	service/roommanager.go:1003	created TURN password	{"username": "MV1QplXNwtUeiDugeVQQxndNVXYzMkaClGTTBVSQFE", "password": "9h0RrfottDvMa5p0JB6BnWEKo5gcnPnuVKr8QJh0x8b"}
2026-01-04T13:34:04.722Z	INFO	livekit	rtc/room.go:1249	participant active	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "agent-AJ_cAurGAkEUra2", "pID": "PA_ApqyKp5yiCUL", "remote": false, "subscriberCandidates": ["[local][selected:1][trickle] udp4 host 100.107.186.2:50085 (resolved: 100.107.186.2:50085)", "[local][trickle] tcp4 host 100.107.186.2:7881 (resolved: 100.107.186.2:7881)", "[remote][selected:1][trickle] udp4 host 192.168.144.4:36114"], "connectionType": "udp", "clientInfo": {"sdk": "PYTHON", "version": "1.0.19", "protocol": 16}, "connectTime": "166.02971ms"}
2026-01-04T13:40:36.384Z	INFO	livekit	rtc/participant.go:1366	participant closing	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "voice_assistant_user_3923", "pID": "PA_gMjTYm6w8eZX", "remote": false, "sendLeave": true, "reason": "CLIENT_REQUEST_LEAVE", "isExpectedToResume": false}
2026-01-04T13:40:36.384Z	WARN	livekit.transport	rtc/transport.go:919	error reading data channel	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "voice_assistant_user_3923", "pID": "PA_gMjTYm6w8eZX", "remote": false, "transport": "PUBLISHER", "label": "_lossy", "error": "dtls timeout: read/write timeout: context deadline exceeded"}
2026-01-04T13:40:36.384Z	WARN	livekit.transport	rtc/transport.go:919	error reading data channel	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "voice_assistant_user_3923", "pID": "PA_gMjTYm6w8eZX", "remote": false, "transport": "PUBLISHER", "label": "_reliable", "error": "dtls timeout: read/write timeout: context deadline exceeded"}
2026-01-04T13:40:36.384Z	WARN	livekit.transport	rtc/transport.go:1276	error reading data channel	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "voice_assistant_user_3923", "pID": "PA_gMjTYm6w8eZX", "remote": false, "transport": "SUBSCRIBER", "label": "_reliable", "error": "dtls timeout: read/write timeout: context deadline exceeded"}
2026-01-04T13:40:36.384Z	WARN	livekit.transport	rtc/transport.go:1276	error reading data channel	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "voice_assistant_user_3923", "pID": "PA_gMjTYm6w8eZX", "remote": false, "transport": "SUBSCRIBER", "label": "_lossy", "error": "dtls timeout: read/write timeout: context deadline exceeded"}
2026-01-04T13:40:37.202Z	INFO	livekit.room	rtc/room.go:807	closing room	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9"}
2026-01-04T13:40:37.202Z	INFO	livekit	rtc/participant.go:1366	participant closing	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "agent-AJ_cAurGAkEUra2", "pID": "PA_ApqyKp5yiCUL", "remote": false, "sendLeave": true, "reason": "ROOM_CLOSED", "isExpectedToResume": false}
2026-01-04T13:40:37.202Z	INFO	livekit.agents	agent/worker.go:530	job ended	{"workerID": "AW_TKMenpwESCKQ", "agentName": "", "jobType": "JT_ROOM", "jobID": "AJ_cAurGAkEUra2", "status": "JS_FAILED", "error": "agent worker left the room"}
2026-01-04T13:40:37.203Z	WARN	livekit.transport	rtc/transport.go:1276	error reading data channel	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "agent-AJ_cAurGAkEUra2", "pID": "PA_ApqyKp5yiCUL", "remote": false, "transport": "SUBSCRIBER", "label": "_reliable", "error": "dtls timeout: read/write timeout: context deadline exceeded"}
2026-01-04T13:40:37.203Z	WARN	livekit.transport	rtc/transport.go:1276	error reading data channel	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "participant": "agent-AJ_cAurGAkEUra2", "pID": "PA_ApqyKp5yiCUL", "remote": false, "transport": "SUBSCRIBER", "label": "_lossy", "error": "dtls timeout: read/write timeout: context deadline exceeded"}
2026-01-04T13:40:37.203Z	INFO	livekit	service/roommanager.go:187	deleting room state	{"room": "voice_assistant_room"}
2026-01-04T13:40:37.203Z	INFO	livekit.room	service/roommanager.go:664	room closed	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9"}
2026-01-04T13:40:37.203Z	INFO	livekit.room	rtc/room.go:791	closing idle room	{"room": "voice_assistant_room", "roomID": "RM_GhdCo264nka9", "reason": "departure timeout"}

I think the problem is somewhere in the voice_agent.py - it's not initializing mcp servers properly (I use only the default ones). I've increased the logging level to debug, added several log messages, but still no clue why it's not getting past the load_mcp_config() call.

Here are the caal-agent logs:

nnxruntime cpuid_info warning: Unknown CPU vendor. cpuinfo_vendor value: 0
Preloading models...
  Loading STT: mlx-community/whisper-large-v3-turbo
Starting new HTTP connection (1): host.docker.internal:8001
http://host.docker.internal:8001 "POST /v1/models?model_name=mlx-community/whisper-large-v3-turbo HTTP/1.1" 200 94
  ✓ STT ready
  Loading LLM: ministral-3:8b (num_ctx=8192)
Starting new HTTP connection (1): host.docker.internal:11434
http://host.docker.internal:11434 "POST /api/generate HTTP/1.1" 200 None
  ✓ LLM ready
starting worker
preloading plugins
INFO starting worker | name=livekit.agents version=1.3.3 rtc-version=1.0.19 timestamp=2026-01-04T13:33:48.301667+00:00
INFO preloading plugins | name=livekit.agents packages=["livekit.plugins.silero","livekit.plugins.openai","av"] timestamp=2026-01-04T13:33:48.302089+00:00
onnxruntime cpuid_info warning: Unknown CPU vendor. cpuinfo_vendor value: 0
initializing process
INFO initializing process | name=livekit.agents pid=53 timestamp=2026-01-04T13:33:49.655008+00:00
initializing process
INFO initializing process | name=livekit.agents pid=55 timestamp=2026-01-04T13:33:49.656460+00:00
INFO initializing process | name=livekit.agents pid=57 timestamp=2026-01-04T13:33:49.658347+00:00
initializing process
initializing process
INFO initializing process | name=livekit.agents pid=59 timestamp=2026-01-04T13:33:49.660095+00:00
Using selector: EpollSelector
Using selector: EpollSelector
Using selector: EpollSelector
Using selector: EpollSelector
process initialized
INFO process initialized | name=livekit.agents pid=53 elapsed_time=1.5 timestamp=2026-01-04T13:33:51.154807+00:00
process initialized
INFO process initialized | name=livekit.agents pid=55 elapsed_time=1.5 timestamp=2026-01-04T13:33:51.154956+00:00
process initialized
INFO process initialized | name=livekit.agents pid=57 elapsed_time=1.5 timestamp=2026-01-04T13:33:51.155012+00:00
process initialized
INFO process initialized | name=livekit.agents pid=59 elapsed_time=1.5 timestamp=2026-01-04T13:33:51.155197+00:00
registered worker
INFO registered worker | name=livekit.agents agent_name= id=AW_TKMenpwESCKQ url=ws://livekit:7880 region= protocol=16 timestamp=2026-01-04T13:33:51.160787+00:00
received job request
INFO received job request | name=livekit.agents job_id=AJ_cAurGAkEUra2 dispatch_id=AD_JNXxcHP7JeGF room=voice_assistant_room room_id=RM_GhdCo264nka9 agent_name= resuming=false enable_recording=false timestamp=2026-01-04T13:34:04.528849+00:00
Joining room: voice_assistant_room
Joining room: voice_assistant_room
DEBUG Joining room: voice_assistant_room | name=voice-agent pid=53 job_id=AJ_cAurGAkEUra2 room_id=RM_GhdCo264nka9 timestamp=2026-01-04T13:34:04.545973+00:00
initializing process
INFO initializing process | name=livekit.agents pid=123 timestamp=2026-01-04T13:34:04.588681+00:00
Starting webhook server on port 8889
joined room
Loaded MCP server config: n8n (http://host.docker.internal:5678/mcp-server/http)
Loaded settings from /app/settings.json
INFO process initialized | name=livekit.agents pid=123 elapsed_time=1.2 timestamp=2026-01-04T13:34:05.785324+00:00
process initialized
Using selector: EpollSelector
INFO process exiting | name=livekit.agents reason= pid=53 job_id=AJ_cAurGAkEUra2 room_id=RM_GhdCo264nka9 timestamp=2026-01-04T13:40:37.203099+00:00
process exiting
received job request
INFO received job request | name=livekit.agents job_id=AJ_hJtDY23pLN6K dispatch_id=AD_AQ3LNa6tawqf room=voice_assistant_room room_id=RM_jD8RTiCizhkA agent_name= resuming=false enable_recording=false timestamp=2026-01-04T13:47:39.997224+00:00
Joining room: voice_assistant_room
Joining room: voice_assistant_room
DEBUG Joining room: voice_assistant_room | name=voice-agent pid=55 job_id=AJ_hJtDY23pLN6K room_id=RM_jD8RTiCizhkA timestamp=2026-01-04T13:47:40.006184+00:00
initializing process
INFO initializing process | name=livekit.agents pid=158 timestamp=2026-01-04T13:47:40.059652+00:00
Starting webhook server on port 8889
joined room
Loaded MCP server config: n8n (http://host.docker.internal:5678/mcp-server/http)
Loaded settings from /app/settings.json
INFO process initialized | name=livekit.agents pid=158 elapsed_time=1.08 timestamp=2026-01-04T13:47:41.136321+00:00
process initialized
Using selector: EpollSelector
ignoring text stream with topic 'lk.chat', no callback attached
process exiting
INFO process exiting | name=livekit.agents reason= pid=55 job_id=AJ_hJtDY23pLN6K room_id=RM_jD8RTiCizhkA timestamp=2026-01-04T13:48:05.206072+00:00
received job request
INFO received job request | name=livekit.agents job_id=AJ_AK8r4uQdAUSA dispatch_id=AD_bbkMrauxWVUG room=voice_assistant_room room_id=RM_HcKBeDTkmhaX agent_name= resuming=false enable_recording=false timestamp=2026-01-04T13:50:24.335292+00:00
Joining room: voice_assistant_room
Joining room: voice_assistant_room
DEBUG Joining room: voice_assistant_room | name=voice-agent pid=57 job_id=AJ_AK8r4uQdAUSA room_id=RM_HcKBeDTkmhaX timestamp=2026-01-04T13:50:24.344464+00:00
initializing process
INFO initializing process | name=livekit.agents pid=193 timestamp=2026-01-04T13:50:24.418002+00:00
Starting webhook server on port 8889
joined room
Loaded MCP server config: n8n (http://host.docker.internal:5678/mcp-server/http)
Loaded settings from /app/settings.json
INFO process initialized | name=livekit.agents pid=193 elapsed_time=1.09 timestamp=2026-01-04T13:50:25.508291+00:00
process initialized
Using selector: EpollSelector
Announce failed: no session in room voice_assistant_room

[mmaudet]

Hi @HubertReX,

Thanks for reporting this issue! I've been working on a fix and I believe I've identified the root causes.

What was wrong

The main issue was missing mlx-audio dependencies for Apple Silicon. The mlx-audio package requires many additional dependencies that aren't automatically installed:

• For Whisper STT: numba, tiktoken, scipy, tqdm
• For Kokoro TTS: loguru, misaki, num2words, spacy, phonemizer-fork, espeakng-loader, torch

Without these, the models fail to load silently, causing the agent to not respond.

Additionally:

• Python 3.12+ has compatibility issues with pydantic (Python 3.11 is required)
• Some error handling improvements were needed in the voice agent

The fix

I've submitted a PR (#12) that includes:

1. Improved start-apple.sh: Now automatically creates a dedicated virtual environment (~/.mlx-audio-venv) with Python 3.11 and installs all required dependencies
2. Updated documentation: Complete dependency list in README-APPLE.md
3. Better error handling: Timeout handling and race condition fixes in voice_agent.py

Can you test it?

To help validate the fix before merge, could you try a fresh installation from my fork?

# Remove existing installation
rm -rf ~/.mlx-audio-venv
docker compose -f docker-compose.apple.yaml down --rmi all

# Clone from fork with fix
git clone -b fix/voice-or-chat-not-working-apple-docker https://github.com/mmaudet/CAAL.git caal-test
cd caal-test
cp .env.example .env
# Edit .env with your settings (CAAL_HOST_IP, etc.)

# Run the improved script
./start-apple.sh

The script will now automatically set up everything. Please let me know if voice and chat work for you!

Thanks for your patience.

[HubertReX]

@mmaudet thanks for a quick response and detailed description 👍

I've done a fresh installation step by step as per your instructions, but unfortunately, the result is the same - the agent is waiting for a command, but there is no response.

I'll try to debug it on my own in the next couple of days. Perhaps I'll figure it out.

Meanwhile, I have some minor suggestions:

1. MLX_VENV location (start-apple.sh): I'd rather keep it inside the project folder (MLX_VENV="~/.mlx-audio-venv" => MLX_VENV=".mlx-audio-venv")
2. Whisper preloading is wrong (voice_agent.py): I believe this line should be changed from f"{speaches_url}/v1/models/{whisper_model}", to f"{speaches_url}/v1/models?model_name={whisper_model}",) otherwise in logs you see STT model download returned 404
3. the progress_bar function (start-apple.sh): It's not used
4. Handle https docker profile (start-apple.sh): It would be nice to add something like this to the start script:

if [ -n "${HTTPS_DOMAIN}" ]; then
  DOCKER_PROFILE='--profile https'
else
  DOCKER_PROFILE=''
fi

so then each docker compose command could look like this:

docker compose -f docker-compose.apple.yaml "$DOCKER_PROFILE" down
docker compose -f docker-compose.apple.yaml "$DOCKER_PROFILE" up -d

and this will work with and without https.

[mmaudet]

Hi @HubertReX,

Thanks for testing and for your detailed feedback! We've just pushed an update to the fix branch that incorporates all your suggestions:

Changes in start-apple.sh:

• ✅ Moved MLX_VENV from ~/.mlx-audio-venv to .mlx-audio-venv (project-local)
• ✅ Removed unused progress_bar function
• ✅ Added automatic .env loading and conditional HTTPS profile support (--profile https when HTTPS_DOMAIN is set)

Also fixed in voice_agent.py:

• ✅ Corrected Whisper model URL from /v1/models/{model} to /v1/models?model_name={model}

---

The symptoms you describe (DTLS timeout, JS_FAILED, no session) suggest the agent may be crashing during initialization before it can register its session. Could you pull the latest changes and run these diagnostic commands?

# Pull latest fix
git fetch origin
git checkout fix/voice-or-chat-not-working-apple-docker
git pull origin fix/voice-or-chat-not-working-apple-docker

# Clean previous venv (now project-local)
rm -rf .mlx-audio-venv

# Fresh start
./start-apple.sh --stop
./start-apple.sh

If it still doesn't work, these diagnostics would help identify the issue:

1. Check environment configuration

# Verify CAAL_HOST_IP is set correctly
grep CAAL_HOST_IP .env
# Should output your Mac's local IP (e.g., 192.168.1.x), NOT localhost or 127.0.0.1

# Get your actual IP for comparison
ipconfig getifaddr en0

2. Test connectivity from Docker to host services

# Test mlx-audio connectivity from inside Docker
docker run --rm --add-host=host.docker.internal:host-gateway alpine/curl \
  curl -s http://host.docker.internal:8001/docs | head -5

# Test Ollama connectivity  
docker run --rm --add-host=host.docker.internal:host-gateway alpine/curl \
  curl -s http://host.docker.internal:11434/api/tags | head -5

3. Check agent startup logs (verbose)

# Watch agent logs for errors
docker compose -f docker-compose.apple.yaml logs -f agent 2>&1 | head -100

4. Check mlx-audio and health status

# mlx-audio running?
curl -s http://localhost:8001/docs > /dev/null && echo "✓ mlx-audio OK" || echo "✗ mlx-audio not responding"

# Agent health endpoint
curl -s http://localhost:8889/health | python3 -m json.tool

The most common issue on Apple Silicon is CAAL_HOST_IP not being set to the actual LAN IP. Make sure your .env has:

CAAL_HOST_IP=<your-mac-ip>  # Result of: ipconfig getifaddr en0

Let me know what output you get — we're happy to help debug further!

[HubertReX]

@mmaudet thanks again for your support 🙏

I've actually found a problem on my side! It was an issue with N8N_MCP_URL. I used http instead of https(https://<my_domain>.ts.net/mcp-server/http). It's funny, since there are try/except blocks all over the code, but the error is never caught - as you said, the whole process is crashing during initialization, hence there was no error log to pinpoint the source of the error.

Anyway, I'm only a few steps further - now, the TTS is failing with internal server error:

Initialized MCP server: n8n
INFO process initialized | name = livekit.agents pid = 120 elapsed_time = 1.08 timestamp = 2026 - 01 - 04T20: 42: 15.605604 + 00: 00
process initialized
Using selector: EpollSelector
Loading 3 n8n workflows:
  ✓ calendar_get_events
  ✓ youtube_transcript_extractor
  ✓ smart_n8n_backups_with_ollama
== == == == == == == == == == == == == == == == == == == == == == == == == == == == == ==
STARTING VOICE AGENT
== == == == == == == == == == == == == == == == == == == == == == == == == == == == == ==
  STT: http: // host.docker.internal: 8001 (mlx - community / whisper - large - v3 - turbo)
  TTS: http: // host.docker.internal: 8001 (am_puck)
  LLM: Ollama(ministral - 3: 8b, think=False, num_ctx=8192)
  MCP: ['n8n']
== == == == == == == == == == == == == == == == == == == == == == == == == == == == == ==
  Wake word: disabled
  STT instance type: STT
  STT capabilities: streaming = False
  Session STT: STT
failed to synthesize speech, retrying in 0.1s
Traceback(most recent call last):
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/tts.py", line 265, in _main_task
    await self._run(output_emitter)
  File "/app/.venv/lib/python3.11/site-packages/livekit/plugins/openai/tts.py", line 238, in _run
    raise APIStatusError(
livekit.agents._exceptions.APIStatusError: Internal Server Error(status_code=500, request_id=None, body=Internal Server Error, retryable=True)
failed to synthesize speech, retrying in 2.0s
Traceback(most recent call last):
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/tts.py", line 265, in _main_task
    await self._run(output_emitter)
  File "/app/.venv/lib/python3.11/site-packages/livekit/plugins/openai/tts.py", line 238, in _run
    raise APIStatusError(
livekit.agents._exceptions.APIStatusError: Internal Server Error(status_code=500, request_id=None, body=Internal Server Error, retryable=True)
failed to synthesize speech, retrying in 2.0s
Traceback(most recent call last):
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/tts.py", line 265, in _main_task
    await self._run(output_emitter)
  File "/app/.venv/lib/python3.11/site-packages/livekit/plugins/openai/tts.py", line 238, in _run
    raise APIStatusError(
livekit.agents._exceptions.APIStatusError: Internal Server Error(status_code=500, request_id=None, body=Internal Server Error, retryable=True)
Error in _tts_inference_task
Traceback(most recent call last):
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/utils/log.py", line 16, in async_fn_logs
    return await fn(*args, **kwargs)
           ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/voice/generation.py", line 288, in _tts_inference_task
    pushed_duration += await _tts_node_inference(input_segment, pushed_duration)
                       ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^
  File "/app/.venv/lib/python3.11/site-packages/opentelemetry/util/_decorator.py", line 71, in async_wrapper
    return await func(*args, **kwargs)  # type: ignore
           ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/voice/generation.py", line 246, in _tts_node_inference
    async for audio_frame in tts_node:
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/voice/agent.py", line 446, in tts_node
    async for ev in stream:
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/tts.py", line 572, in __anext__
    raise exc  # noqa: B904
    ^ ^ ^ ^ ^ ^ ^ ^ ^
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/tts.py", line 362, in _traceable_main_task
    await self._main_task()
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/tts.py", line 396, in _main_task
    await self._run(output_emitter)
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/stream_adapter.py", line 144, in _run
    await asyncio.gather(*tasks)
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/stream_adapter.py", line 134, in _synthesize
    async for audio in tts_stream:
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/tts.py", line 325, in __anext__
    raise exc  # noqa: B904
    ^ ^ ^ ^ ^ ^ ^ ^ ^
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/tts.py", line 180, in _traceable_main_task
    await self._main_task()
  File "/app/.venv/lib/python3.11/site-packages/livekit/agents/tts/tts.py", line 265, in _main_task
    await self._run(output_emitter)
  File "/app/.venv/lib/python3.11/site-packages/livekit/plugins/openai/tts.py", line 238, in _run
    raise APIStatusError(
livekit.agents._exceptions.APIStatusError: Internal Server Error(status_code=500, request_id=None, body=Internal Server Error, retryable=True)
Agent ready - listening for speech...

I'll keep digging further.

[mmaudet]

Hi @HubertReX,

Great progress on finding the N8N_MCP_URL issue!

The TTS 500 errors are caused by an incompatible voice name. The default voice am_puck exists in kokoro-fastapi (NVIDIA) but not in mlx-audio Kokoro-82M (Apple Silicon).

Quick Fix

Edit your .env and change the TTS voice to one that exists in mlx-audio:

# Change this:
TTS_VOICE=am_puck

# To this:
TTS_VOICE=af_heart

Then restart:

./start-apple.sh --stop
./start-apple.sh

Available voices in mlx-audio Kokoro-82M

The voice format is [lang][gender]_[name]:

• af_heart - American female (recommended default)
• af_bella - American female
• af_nova - American female
• af_alloy - American female
• bf_emma - British female

You can also test TTS directly:

curl -X POST "http://localhost:8001/v1/audio/speech" \
  -H "Content-Type: application/json" \
  -d '{"model": "prince-canuma/Kokoro-82M", "input": "Hello, this is a test.", "voice": "af_heart"}' \
  --output test.mp3 && afplay test.mp3

Let me know if this resolves the TTS errors!

[HubertReX]

The curl -X POST "http://localhost:8001/v1/audio/speech" from above results with this error:

INFO:     127.0.0.1:64602 - "POST /v1/audio/speech HTTP/1.1" 200 OK
2026-01-04 22:14:21.500 | INFO     | mlx_audio.tts.models.kokoro.kokoro:_get_pipeline:254 - Creating new KokoroPipeline for language: a
/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/bin/python: No module named pip
ERROR:    Traceback (most recent call last):
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/runners.py", line 195, in run
    return runner.run(main)
           ^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/runners.py", line 118, in run
    return self._loop.run_until_complete(task)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/base_events.py", line 678, in run_until_complete
    self.run_forever()
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/base_events.py", line 645, in run_forever
    self._run_once()
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/base_events.py", line 1999, in _run_once
    handle._run()
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/events.py", line 88, in _run
    self._context.run(self._callback, *self._args)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/responses.py", line 273, in wrap
    await func()
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/responses.py", line 253, in stream_response
    async for chunk in self.body_iterator:
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/server.py", line 228, in generate_audio
    for result in model.generate(
                  ^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/tts/models/kokoro/kokoro.py", line 271, in generate
    pipeline = self._get_pipeline(lang_code)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/tts/models/kokoro/kokoro.py", line 255, in _get_pipeline
    self._pipelines[lang_code] = KokoroPipeline(
                                 ^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/tts/models/kokoro/pipeline.py", line 99, in __init__
    self.g2p = en.G2P(
               ^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/misaki/en.py", line 501, in __init__
    spacy.cli.download(name)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/spacy/cli/download.py", line 92, in download
    download_model(filename, pip_args, custom_url)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/spacy/cli/download.py", line 180, in download_model
    run_command(cmd)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/spacy/util.py", line 1093, in run_command
    sys.exit(ret.returncode)
SystemExit: 1

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/routing.py", line 701, in lifespan
    await receive()
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/uvicorn/lifespan/on.py", line 137, in receive
    return await self.receive_queue.get()
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/queues.py", line 158, in get
    await getter
asyncio.exceptions.CancelledError

ERROR:    Exception in ASGI application
  + Exception Group Traceback (most recent call last):
  |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/_utils.py", line 79, in collapse_excgroups
  |     yield
  |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/responses.py", line 270, in __call__
  |     async with anyio.create_task_group() as task_group:
  |                ^^^^^^^^^^^^^^^^^^^^^^^^^
  |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/anyio/_backends/_asyncio.py", line 783, in __aexit__
  |     raise BaseExceptionGroup(
  | BaseExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
  +-+---------------- 1 ----------------
    | Traceback (most recent call last):
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/uvicorn/protocols/http/h11_impl.py", line 410, in run_asgi
    |     result = await app(  # type: ignore[func-returns-value]
    |              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/uvicorn/middleware/proxy_headers.py", line 60, in __call__
    |     return await self.app(scope, receive, send)
    |            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/fastapi/applications.py", line 1135, in __call__
    |     await super().__call__(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/applications.py", line 107, in __call__
    |     await self.middleware_stack(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/middleware/errors.py", line 164, in __call__
    |     await self.app(scope, receive, _send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/middleware/cors.py", line 85, in __call__
    |     await self.app(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/middleware/exceptions.py", line 63, in __call__
    |     await wrap_app_handling_exceptions(self.app, conn)(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app
    |     await app(scope, receive, sender)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/fastapi/middleware/asyncexitstack.py", line 18, in __call__
    |     await self.app(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/routing.py", line 716, in __call__
    |     await self.middleware_stack(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/routing.py", line 736, in app
    |     await route.handle(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/routing.py", line 290, in handle
    |     await self.app(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/fastapi/routing.py", line 115, in app
    |     await wrap_app_handling_exceptions(app, request)(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app
    |     await app(scope, receive, sender)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/fastapi/routing.py", line 102, in app
    |     await response(scope, receive, send)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/responses.py", line 269, in __call__
    |     with collapse_excgroups():
    |          ^^^^^^^^^^^^^^^^^^^^
    |   File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/contextlib.py", line 158, in __exit__
    |     self.gen.throw(value)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/_utils.py", line 85, in collapse_excgroups
    |     raise exc
    |   File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/runners.py", line 195, in run
    |     return runner.run(main)
    |            ^^^^^^^^^^^^^^^^
    |   File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/runners.py", line 118, in run
    |     return self._loop.run_until_complete(task)
    |            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |   File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/base_events.py", line 678, in run_until_complete
    |     self.run_forever()
    |   File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/base_events.py", line 645, in run_forever
    |     self._run_once()
    |   File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/base_events.py", line 1999, in _run_once
    |     handle._run()
    |   File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/events.py", line 88, in _run
    |     self._context.run(self._callback, *self._args)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/responses.py", line 273, in wrap
    |     await func()
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/responses.py", line 253, in stream_response
    |     async for chunk in self.body_iterator:
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/server.py", line 228, in generate_audio
    |     for result in model.generate(
    |                   ^^^^^^^^^^^^^^^
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/tts/models/kokoro/kokoro.py", line 271, in generate
    |     pipeline = self._get_pipeline(lang_code)
    |                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/tts/models/kokoro/kokoro.py", line 255, in _get_pipeline
    |     self._pipelines[lang_code] = KokoroPipeline(
    |                                  ^^^^^^^^^^^^^^^
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/tts/models/kokoro/pipeline.py", line 99, in __init__
    |     self.g2p = en.G2P(
    |                ^^^^^^^
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/misaki/en.py", line 501, in __init__
    |     spacy.cli.download(name)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/spacy/cli/download.py", line 92, in download
    |     download_model(filename, pip_args, custom_url)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/spacy/cli/download.py", line 180, in download_model
    |     run_command(cmd)
    |   File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/spacy/util.py", line 1093, in run_command
    |     sys.exit(ret.returncode)
    | SystemExit: 1
    +------------------------------------

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/uvicorn/protocols/http/h11_impl.py", line 410, in run_asgi
    result = await app(  # type: ignore[func-returns-value]
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/uvicorn/middleware/proxy_headers.py", line 60, in __call__
    return await self.app(scope, receive, send)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/fastapi/applications.py", line 1135, in __call__
    await super().__call__(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/applications.py", line 107, in __call__
    await self.middleware_stack(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/middleware/errors.py", line 164, in __call__
    await self.app(scope, receive, _send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/middleware/cors.py", line 85, in __call__
    await self.app(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/middleware/exceptions.py", line 63, in __call__
    await wrap_app_handling_exceptions(self.app, conn)(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app
    await app(scope, receive, sender)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/fastapi/middleware/asyncexitstack.py", line 18, in __call__
    await self.app(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/routing.py", line 716, in __call__
    await self.middleware_stack(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/routing.py", line 736, in app
    await route.handle(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/routing.py", line 290, in handle
    await self.app(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/fastapi/routing.py", line 115, in app
    await wrap_app_handling_exceptions(app, request)(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app
    await app(scope, receive, sender)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/fastapi/routing.py", line 102, in app
    await response(scope, receive, send)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/responses.py", line 269, in __call__
    with collapse_excgroups():
         ^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/contextlib.py", line 158, in __exit__
    self.gen.throw(value)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/_utils.py", line 85, in collapse_excgroups
    raise exc
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/runners.py", line 195, in run
    return runner.run(main)
           ^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/runners.py", line 118, in run
    return self._loop.run_until_complete(task)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/base_events.py", line 678, in run_until_complete
    self.run_forever()
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/base_events.py", line 645, in run_forever
    self._run_once()
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/base_events.py", line 1999, in _run_once
    handle._run()
  File "/opt/homebrew/Cellar/python@3.12/3.12.12/Frameworks/Python.framework/Versions/3.12/lib/python3.12/asyncio/events.py", line 88, in _run
    self._context.run(self._callback, *self._args)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/responses.py", line 273, in wrap
    await func()
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/starlette/responses.py", line 253, in stream_response
    async for chunk in self.body_iterator:
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/server.py", line 228, in generate_audio
    for result in model.generate(
                  ^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/tts/models/kokoro/kokoro.py", line 271, in generate
    pipeline = self._get_pipeline(lang_code)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/tts/models/kokoro/kokoro.py", line 255, in _get_pipeline
    self._pipelines[lang_code] = KokoroPipeline(
                                 ^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/mlx_audio/tts/models/kokoro/pipeline.py", line 99, in __init__
    self.g2p = en.G2P(
               ^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/misaki/en.py", line 501, in __init__
    spacy.cli.download(name)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/spacy/cli/download.py", line 92, in download
    download_model(filename, pip_args, custom_url)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/spacy/cli/download.py", line 180, in download_model
    run_command(cmd)
  File "/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/spacy/util.py", line 1093, in run_command
    sys.exit(ret.returncode)
SystemExit: 1
INFO:     Waiting for child process [57410]
INFO:     Child process [57410] died
/Users/hubertnafalski/Documents/Projects/mlx-audio/.venv/lib/python3.12/site-packages/webrtcvad.py:1: UserWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html. The pkg_resources package is slated for removal as early as 2025-11-30. Refrain from using this package or pin to Setuptools<81.
  import pkg_resources
INFO:     Started server process [58333]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
^CINFO:     Shutting down
INFO:     Shutting down
INFO:     Waiting for application shutdown.
INFO:     Application shutdown complete.
INFO:     Finished server process [58333]
INFO:     Waiting for application shutdown.
INFO:     Application shutdown complete.
INFO:     Finished server process [57923]
INFO:     Received SIGINT, exiting.
INFO:     Waiting for child process [58333]
INFO:     Waiting for child process [57923]
INFO:     Stopping parent process [57408]

[mmaudet]

Hi @HubertReX,

Thanks for the update! We are on the track of the success ;-)

I think the the error No module named pip indicates your virtual environment is corrupted or you're using a different one than what start-apple.sh creates.

The Issue

The path /Users/.../mlx-audio/.venv/ suggests you have a manually created venv. The start-apple.sh script now creates its venv at .mlx-audio-venv (in the project directory).

Clean Reinstall

Please do a complete cleanup and fresh install:

# 1. Stop everything
./start-apple.sh --stop

# 2. Remove ALL old mlx-audio virtual environments
rm -rf .mlx-audio-venv           # New location (project-local)
rm -rf ~/.mlx-audio-venv         # Old location (home directory)
rm -rf ~/mlx-audio/.venv         # If you created one manually

# 3. Pull latest changes (includes TTS voice fix)
git pull origin fix/voice-or-chat-not-working-apple-docker

# 4. Fresh start - this will recreate the venv with all dependencies
./start-apple.sh

The script will:

1. Create a fresh venv at .mlx-audio-venv
2. Install pip and all dependencies (mlx-audio, spacy, misaki, phonemizer-fork, etc.)
3. Download Whisper and Kokoro models
4. Start the mlx-audio server
5. Start Docker services

Verify Installation

After startup, test TTS directly:

# Test TTS with the correct voice
curl -X POST "http://localhost:8001/v1/audio/speech" \
  -H "Content-Type: application/json" \
  -d '{"model": "prince-canuma/Kokoro-82M", "input": "Hello, this is a test.", "voice": "af_heart"}' \
  --output test.mp3 && afplay test.mp3

If you still see errors, check the mlx-audio logs:

tail -f /tmp/caal-mlx-audio.log

Let me know how it goes!

[HubertReX]

Yupi 🎉🎊🥳 It works!
You were right - it was venv issue. I couldn't get to work with .mlx-audio-venv (python 3.11) - got some funny errors - the library wanted to use a part of my user name (my first name) from the path as model name 🤔 Gemini claims that this is all because it's similar to BERT model 🤣

tail -f /tmp/caal-mlx-audio.log
                              ^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/caal-test/.mlx-audio-venv/lib/python3.11/site-packages/mlx_audio/utils.py", line 138, in load_model
    return model_loaders[model_category](model_name)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/caal-test/.mlx-audio-venv/lib/python3.11/site-packages/mlx_audio/tts/utils.py", line 241, in load_model
    model_class, model_type = get_model_and_args(
                              ^^^^^^^^^^^^^^^^^^^
  File "/Users/hubertnafalski/Documents/Projects/caal-test/.mlx-audio-venv/lib/python3.11/site-packages/mlx_audio/tts/utils.py", line 127, in get_model_and_args
    raise ValueError(msg)
ValueError: Model type hubert not supported.

I had a separate installation of mlx-audio, outside this project, with python 3.12 where this error didn't show up, but I've created it using uv. Venv created doesn't have pip, but 'mlx-audioneeds it to install models. After addingpip` to this venv, all worked like a charm 👌

Once again, thanks for your help and patience - all in all, generally the project instructions are all good - I think I had some corner case errors.

[mmaudet]

Hi @HubertReX,

Great to hear everything is working now! 🎉

Thanks for your patience and detailed feedback throughout this debugging process. Your findings were invaluable:

1. HTTP vs HTTPS for N8N_MCP_URL - Silent crash during initialization
2. TTS voice incompatibility - am_puck doesn't exist in mlx-audio Kokoro
3. uv-created venvs missing pip - Caused "No module named pip" errors

Based on your last finding, we've just pushed another improvement to start-apple.sh:

# The script now:
# 1. Detects corrupted venvs (missing pip) and recreates them automatically
# 2. Verifies pip is available after venv creation
# 3. Checks for pip in the initial health check

This means future users who accidentally create a venv with uv or have a corrupted environment will get automatic recovery instead of cryptic errors.

Thanks again for being such a thorough beta tester! Your contributions have significantly improved the Apple Silicon experience. 🙏

---

Summary of fixes in this branch:

• ✅ Improved start-apple.sh with auto .env loading and HTTPS profile support
• ✅ Fixed Whisper model URL (/v1/models?model_name=)
• ✅ Default TTS voice set to af_heart for mlx-audio compatibility
• ✅ Automatic detection and recovery of corrupted venvs

[cmac86]

Thanks to both of you for working through this problem. My main setup is not based around mlx-audio on the Mac so this is very helpful. @mmaudet thank you for the support! I will get the PR merged.