Bump to version 3.6.0

The agent will now only connect to Rooms when explicitly requested
via the Agent Dispatch API instead of auto-connecting to all new Rooms.

community/.env

@@ -30,7 +30,5 @@ MINIO_SECRET_KEY=minioadmin
 MONGO_ADMIN_USERNAME=mongoadmin
 MONGO_ADMIN_PASSWORD=mongoadmin
 
-# OpenVidu Meet configuration.
-MEET_INITIAL_ADMIN_USER=admin
-MEET_INITIAL_ADMIN_PASSWORD=admin
-MEET_INITIAL_API_KEY=meet-api-key
+# OpenVidu Meet base path
+MEET_BASE_PATH=/meet

community/agent-speech-processing.yaml

@@ -1,5 +1,5 @@
 # Docker image of the agent.
-docker_image: docker.io/openvidu/agent-speech-processing:3.4.0
+docker_image: docker.io/openvidu/agent-speech-processing-vosk:3.6.0
 
 # Whether to run the agent or not.
 enabled: false
@@ -7,18 +7,18 @@ enabled: false
 # Maximum CPU load threshold for the agent to accept new jobs. Value between 0 and 1.
 load_threshold: 1.0
 
-# Log level for the agent [DEBUG, INFO, WARNING, ERROR, CRITICAL]
+# Log level for the agent [DEBUG, INFO, WARN, ERROR, CRITICAL]
 log_level: INFO
 
 live_captions:
-  # How this agent will connect to Rooms [automatic, manual]
-  # - automatic: the agent will automatically connect to new Rooms.
+  # How this agent will connect to Rooms [manual, automatic]
   # - manual: the agent will connect to new Rooms only when your application dictates it by using the Agent Dispatch API.
-  processing: automatic
+  # - automatic: the agent will automatically connect to new Rooms.
+  processing: manual
 
-  # Which speech-to-text AI provider to use [aws, azure, google, openai, groq, deepgram, assemblyai, fal, clova, speechmatics, gladia, sarvam, spitch]
+  # Which speech-to-text AI provider to use [aws, azure, google, openai, azure_openai, groq, deepgram, assemblyai, fal, clova, speechmatics, gladia, sarvam, mistralai, cartesia, soniox, nvidia, elevenlabs, simplismart, vosk, sherpa]
   # The custom configuration for the selected provider must be set below
-  provider:
+  provider: vosk
 
   aws:
     # Credentials for AWS Transcribe. See https://docs.aws.amazon.com/transcribe/latest/dg/what-is.html
@@ -63,6 +63,10 @@ live_captions:
     # Removes profanity (swearing), or replaces letters of profane words with stars. Valid values: Masked | Removed | Raw
     # See https://learn.microsoft.com/en-us/azure/ai-services/translator/profanity-filtering
     profanity:
+    # List of words or phrases to boost recognition accuracy. Azure will give higher priority to these phrases during recognition.
+    phrase_list:
+    # Controls punctuation behavior. If True, enables explicit punctuation mode where punctuation marks are added explicitly. If False (default), uses Azure's default punctuation behavior.
+    explicit_punctuation:
 
   azure_openai:
     # Credentials for Azure OpenAI APIs. See https://learn.microsoft.com/en-us/azure/api-management/api-management-authenticate-authorize-azure-openai
@@ -82,6 +86,8 @@ live_captions:
     project:
     # The language code to use for transcription (e.g., "en" for English).
     language:
+    # Whether to automatically detect the language.
+    detect_language:
     # ID of the model to use for speech-to-text.
     model:
     # Initial prompt to guide the transcription.
@@ -135,6 +141,8 @@ live_captions:
     # The language of the input audio. Supplying the input language in ISO-639-1 format
     # (https://en.wikipedia.org/wiki/List_of_ISO_639-1_codes) will improve accuracy and latency.
     language:
+    # Whether to automatically detect the language.
+    detect_language:
     # Optional text prompt to guide the transcription. Only supported for whisper-1.
     prompt:
 
@@ -146,8 +154,12 @@ live_captions:
     # The language of the input audio. Supplying the input language in ISO-639-1 format
     # (https://en.wikipedia.org/wiki/List_of_ISO_639-1_codes) will improve accuracy and latency.
     language:
+    # Whether to automatically detect the language.
+    detect_language:
     # Prompt to guide the model's style or specify how to spell unfamiliar words. 224 tokens max.
     prompt:
+    # Base URL for the Groq API. By default "https://api.groq.com/openai/v1"
+    base_url:
 
   deepgram:
     # See https://console.deepgram.com/
@@ -156,25 +168,27 @@ live_captions:
     model:
     # See https://developers.deepgram.com/reference/speech-to-text-api/listen-streaming#request.query.language
     language:
-    # Whether to enable automatic language detection. Defaults to false. See https://developers.deepgram.com/docs/language-detection
+    # Whether to enable automatic language detection. See https://developers.deepgram.com/docs/language-detection
     detect_language: false
-    # Whether to return interim (non-final) transcription results. Defaults to true. See https://developers.deepgram.com/docs/interim-results
+    # Whether to return interim (non-final) transcription results. See https://developers.deepgram.com/docs/interim-results
     interim_results: true
-    # Whether to apply smart formatting to numbers, dates, etc. Defaults to false. See https://developers.deepgram.com/docs/smart-format
+    # Whether to apply smart formatting to numbers, dates, etc. See https://developers.deepgram.com/docs/smart-format
     smart_format: false
-    # When smart_format is used, ensures it does not wait for sequence to be complete before returning results. Defaults to true. See https://developers.deepgram.com/docs/smart-format#using-no-delay
+    # When smart_format is used, ensures it does not wait for sequence to be complete before returning results. See https://developers.deepgram.com/docs/smart-format#using-no-delay
     no_delay: true
-    # Whether to add punctuations to the transcription. Defaults to true. Turn detector will work better with punctuations. See https://developers.deepgram.com/docs/punctuation
+    # Whether to add punctuations to the transcription. Turn detector will work better with punctuations. See https://developers.deepgram.com/docs/punctuation
     punctuate: true
-    # Whether to include filler words (um, uh, etc.) in transcription. Defaults to true. See https://developers.deepgram.com/docs/filler-words
+    # Whether to include filler words (um, uh, etc.) in transcription. See https://developers.deepgram.com/docs/filler-words
     filler_words: true
-    # Whether to filter profanity from the transcription. Defaults to false. See https://developers.deepgram.com/docs/profanity-filter
+    # Whether to filter profanity from the transcription. See https://developers.deepgram.com/docs/profanity-filter
     profanity_filter: false
-    # List of tuples containing keywords and their boost values for improved recognition. Each tuple should be (keyword: str, boost: float). Defaults to None. keywords does not work with Nova-3 models. Use keyterms instead.
+    # Whether to transcribe numbers as numerals. See https://developers.deepgram.com/docs/numerals
+    numerals: false
+    # List of tuples containing keywords and their boost values for improved recognition. Each tuple should be (keyword: str, boost: float). keywords does not work with Nova-3 models. Use keyterms instead.
     # keywords:
     #   - [OpenVidu, 1.5]
     #   - [WebRTC, 1]
-    # List of key terms to improve recognition accuracy. Defaults to None. keyterms is supported by Nova-3 models.
+    # List of key terms to improve recognition accuracy. keyterms is supported by Nova-3 models.
     # Commented below is an example
     keyterms:
       # - "OpenVidu"
@@ -183,8 +197,18 @@ live_captions:
   assemblyai:
     # API key for AssemblyAI. See https://www.assemblyai.com/dashboard/api-keys
     api_key:
+    # The confidence threshold (0.0 to 1.0) to use when determining if the end of a turn has been reached.
+    end_of_turn_confidence_threshold:
+    # The minimum amount of silence in milliseconds required to detect end of turn when confident.
+    min_end_of_turn_silence_when_confident:
+    # The maximum amount of silence in milliseconds allowed in a turn before end of turn is triggered.
+    max_turn_silence:
     # Whether to return formatted final transcripts (proper punctuation, letter casing...). If enabled, formatted final transcripts are emitted shortly following an end-of-turn detection.
     format_turns: true
+    # List of keyterms to improve recognition accuracy for specific words and phrases.
+    keyterms_prompt:
+      # - "OpenVidu"
+      # - "WebRTC"
 
   fal:
     # API key for fal. See https://fal.ai/dashboard/keys
@@ -208,12 +232,14 @@ live_captions:
   speechmatics:
     # API key for Speechmatics. See https://portal.speechmatics.com/manage-access/
     api_key:
-    # ISO 639-1 language code. All languages are global and can understand different dialects/accents. To see the list of all supported languages, see https://docs.speechmatics.com/introduction/supported-languages
+    # ISO 639-1 language code. All languages are global and can understand different dialects/accents. To see the list of all supported languages, see https://docs.speechmatics.com/speech-to-text/languages#transcription-languages
     language:
-    # Operating point to use for the transcription per required accuracy & complexity. To learn more, see https://docs.speechmatics.com/features/accuracy-language-packs#accuracy
+    # Operating point to use for the transcription per required accuracy & complexity. To learn more, see https://docs.speechmatics.com/speech-to-text/languages#operating-points
     operating_point:
-    # Partial transcripts allow you to receive preliminary transcriptions and update as more context is available until the higher-accuracy final transcript is returned. Partials are returned faster but without any post-processing such as formatting. See https://docs.speechmatics.com/features/realtime-latency#partial-transcripts
+    # Partial transcripts allow you to receive preliminary transcriptions and update as more context is available until the higher-accuracy final transcript is returned. Partials are returned faster but without any post-processing such as formatting. See https://docs.speechmatics.com/speech-to-text/realtime/output#partial-transcripts
     enable_partials:
+    # Enable speaker diarization. When enabled, the STT engine will determine and attribute words to unique speakers. The speaker_sensitivity parameter can be used to adjust the sensitivity of diarization
+    enable_diarization:
     # RFC-5646 language code to make spelling rules more consistent in the transcription output. See https://docs.speechmatics.com/features/word-tagging#output-locale
     output_locale:
     # The delay in seconds between the end of a spoken word and returning the final transcript results. See https://docs.speechmatics.com/features/realtime-latency#configuration-example
@@ -255,6 +281,10 @@ live_captions:
     languages:
     # Whether to allow switching between languages during recognition. Defaults to True
     code_switching:
+    # https://docs.gladia.io/api-reference/v2/live/init#body-pre-processing-audio-enhancer
+    pre_processing_audio_enhancer:
+    # https://docs.gladia.io/api-reference/v2/live/init#body-pre-processing-speech-threshold
+    pre_processing_speech_threshold:
 
   sarvam:
     # API key for Sarvam. See https://dashboard.sarvam.ai/key-management
@@ -263,3 +293,145 @@ live_captions:
     language:
     # The Sarvam STT model to use. See https://docs.sarvam.ai/api-reference-docs/speech-to-text/transcribe#request.body.model.model
     model:
+
+  mistralai:
+    # API key for Mistral AI. See https://console.mistral.ai/api-keys
+    api_key:
+    # Name of the Voxtral STT model to use. Default to voxtral-mini-latest. See https://docs.mistral.ai/capabilities/audio/
+    model:
+    # The language code to use for transcription (e.g., "en" for English)
+    language:
+
+  cartesia:
+    # API key for Cartesia. See https://play.cartesia.ai/keys
+    api_key:
+    # The Cartesia STT model to use
+    model:
+    # The language code to use for transcription (e.g., "en" for English)
+    language:
+
+  soniox:
+    # API key for Soniox. See https://console.soniox.com/
+    api_key:
+    # Set language hints when possible to significantly improve accuracy. See: https://soniox.com/docs/stt/concepts/language-hints
+    language_hints:
+      # - "en"
+      # - "es"
+    # Set context to improve recognition of difficult and rare words. Context is a string and can include words, phrases, sentences, or summaries (limit: 10K chars). See https://soniox.com/docs/stt/concepts/context
+    context:
+
+  nvidia:
+    # API key for NVIDIA. See https://build.nvidia.com/explore/speech?integrate_nim=true&hosted_api=true&modal=integrate-nim
+    # Required when using NVIDIA's cloud services. To use a self-hosted NVIDIA Riva server setup "server" and "use_ssl" instead.
+    api_key:
+    # The NVIDIA Riva ASR model to use. Default is "parakeet-1.1b-en-US-asr-streaming-silero-vad-sortformer"
+    # See available models: https://build.nvidia.com/search/models?filters=usecase%3Ausecase_speech_to_text
+    model:
+    # The NVIDIA function ID for the model. Default is "1598d209-5e27-4d3c-8079-4751568b1081"
+    function_id:
+    # Whether to add punctuation to transcription results. Default is true.
+    punctuate:
+    # The language code for transcription. Default is "en-US"
+    language_code:
+    # Audio sample rate in Hz. Default is 16000.
+    sample_rate:
+    # The NVIDIA Riva server address. Default is "grpc.nvcf.nvidia.com:443"
+    # For self-hosted NIM, use your server address (e.g., "localhost:50051")
+    server:
+    # Whether to use SSL for the connection. Default is true.
+    # Set to false for locally hosted Riva NIM services without SSL.
+    use_ssl:
+
+  spitch:
+    # API key for Spitch. See https://docs.spitch.app/keys
+    api_key:
+    # Language short code for the generated speech. For supported values, see https://docs.spitch.app/concepts/languages
+    language:
+
+  elevenlabs:
+    # API key for ElevenLabs. See https://elevenlabs.io/app/settings/api-keys
+    api_key:
+    # The ElevenLabs STT model to use. Valid values are ["scribe_v1", "scribe_v2", "scribe_v2_realtime"]. See https://elevenlabs.io/docs/overview/models#models-overview
+    model_id:
+    # An ISO-639-1 or ISO-639-3 language_code corresponding to the language of the audio file. Can sometimes improve transcription performance if known beforehand. Defaults to null, in this case the language is predicted automatically
+    language_code:
+    # Custom base URL for the API. Optional.
+    base_url:
+    # Audio sample rate in Hz. Default is 16000.
+    sample_rate:
+    # Whether to tag audio events like (laughter), (footsteps), etc. in the transcription. Only supported for Scribe v1 model. Default is True
+    tag_audio_events:
+    # Whether to include word-level timestamps in the transcription. Default is false.
+    include_timestamps:
+
+  simplismart:
+    # API key for SimpliSmart. See https://docs.simplismart.ai/model-suite/settings/api-keys
+    api_key:
+    # Model identifier for the backend STT model. One of ["openai/whisper-large-v2", "openai/whisper-large-v3", "openai/whisper-large-v3-turbo"]
+    # Default is "openai/whisper-large-v3-turbo"
+    model:
+    # Language code for transcription (default: "en"). See https://docs.simplismart.ai/get-started/playground/transcription-models#supported-languages-with-their-codes
+    language:
+    # Operation to perform. "transcribe" converts speech to text in the original language, "translate" translates into English. Default is "transcribe".
+    task:
+    # If true, disables timestamp generation in transcripts. Default is true
+    without_timestamps:
+    # Minimum duration (ms) for a valid speech segment. Default is 0
+    min_speech_duration_ms:
+    # Decoding temperature (affects randomness). Default is 0.0
+    temperature:
+    # Whether to permit multilingual recognition. Default is false
+    multilingual:
+
+  vosk:
+    # Vosk language model. This provider requires docker_image "docker.io/openvidu/agent-speech-processing-vosk"
+    # Below is the list of pre-installed models in the container (available at https://alphacephei.com/vosk/models):
+    # - vosk-model-en-us-0.22-lgraph (English US)
+    # - vosk-model-small-cn-0.22 (Chinese)
+    # - vosk-model-small-de-0.15 (German)
+    # - vosk-model-small-en-in-0.4 (English India)
+    # - vosk-model-small-es-0.42 (Spanish)
+    # - vosk-model-small-fr-0.22 (French)
+    # - vosk-model-small-hi-0.22 (Hindi)
+    # - vosk-model-small-it-0.22 (Italian)
+    # - vosk-model-small-ja-0.22 (Japanese)
+    # - vosk-model-small-nl-0.22 (Dutch)
+    # - vosk-model-small-pt-0.3 (Portuguese)
+    # - vosk-model-small-ru-0.22 (Russian)
+    model: vosk-model-en-us-0.22-lgraph
+    # Language code for reference. It has no effect other than observability purposes.
+    # If a pre-installed "model" is declared, this will be set automatically if empty.
+    language:
+    # Audio sample rate in Hz. Default is 16000.
+    sample_rate:
+    # Whether to return interim/partial results during recognition. Default is true.
+    partial_results:
+    # Whether to override Vosk's built-in Voice Activity Detection (VAD) with Silero's VAD. Default is false.
+    use_silero_vad: false
+
+  sherpa:
+    # sherpa streaming model. This provider requires docker_image "docker.io/openvidu/agent-speech-processing-sherpa"
+    # Below is the list of pre-installed models in the container (available at https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models):
+    # - sherpa-onnx-streaming-zipformer-en-kroko-2025-08-06 (English)
+    # - sherpa-onnx-streaming-zipformer-es-kroko-2025-08-06 (Spanish)
+    # - sherpa-onnx-streaming-zipformer-de-kroko-2025-08-06 (German)
+    # - sherpa-onnx-streaming-zipformer-fr-kroko-2025-08-06 (French)
+    # - sherpa-onnx-streaming-zipformer-ar_en_id_ja_ru_th_vi_zh-2025-02-10 (Multilingual: Arabic, English, Indonesian, Japanese, Russian, Thai, Vietnamese, Chinese)
+    model: sherpa-onnx-streaming-zipformer-en-kroko-2025-08-06
+    # Language code for reference. Auto-detected from model name if not set.
+    language:
+    # Runtime provider for sherpa-onnx. Supported values: "cpu" or "cuda". Default is "cpu".
+    # Learn about GPU acceleration at https://openvidu.io/docs/ai/live-captions/#gpu-acceleration-for-sherpa-provider
+    provider:
+    # Audio sample rate in Hz. Default is 16000.
+    sample_rate:
+    # Whether to return interim/partial results during recognition. Default is true.
+    partial_results:
+    # Number of threads for ONNX Runtime. Default is 2.
+    num_threads:
+    # Recognizer type ("transducer", "paraformer", "zipformer_ctc", "nemo_ctc", "t_one_ctc"). Auto-detected from model name if not set.
+    recognizer_type:
+    # Decoding method ("greedy_search", "modified_beam_search"). Default is "greedy_search".
+    decoding_method:
+    # Whether to override sherpa's built-in Voice Activity Detection (VAD) with Silero's VAD. Default is false.
+    use_silero_vad: false

community/configure_lan_private_ip_macos.sh

@@ -14,14 +14,4 @@ if [ -z "$LAN_PRIVATE_IP" ]; then
 fi
 
 # Replace the LAN_PRIVATE_IP in the .env file
-sed -i'' -e "s/LAN_PRIVATE_IP=.*/LAN_PRIVATE_IP=$LAN_PRIVATE_IP/g" .env
-
-# If sillicon mac, enable EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU flag
-if [ "$(uname -m)" = "arm64" ]; then
-    if ! grep -q "EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU" .env; then
-        echo "# Enable this flag to run Docker Desktop on Apple Silicon Macs" >> .env
-        echo "EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU=1" >> .env
-    else
-        sed -i'' -e "s/EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU=.*/EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU=1/g" .env
-    fi
-fi
+sed -i'' -e "s/LAN_PRIVATE_IP=.*/LAN_PRIVATE_IP=$LAN_PRIVATE_IP/g" .env

community/docker-compose.yaml

@@ -1,7 +1,6 @@
 services:
   caddy-proxy:
-    image: docker.io/openvidu/openvidu-caddy-local:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-caddy-local:3.6.0
     container_name: caddy-proxy
     restart: unless-stopped
     extra_hosts:
@@ -17,41 +16,40 @@ services:
       - DASHBOARD_ADMIN_PASSWORD=${DASHBOARD_ADMIN_PASSWORD:-}
       - MINIO_ACCESS_KEY=${MINIO_ACCESS_KEY:-}
       - MINIO_SECRET_KEY=${MINIO_SECRET_KEY:-}
-      - MEET_INITIAL_ADMIN_USER=${MEET_INITIAL_ADMIN_USER:-}
-      - MEET_INITIAL_ADMIN_PASSWORD=${MEET_INITIAL_ADMIN_PASSWORD:-}
-      - MEET_INITIAL_API_KEY=${MEET_INITIAL_API_KEY:-}
+      - MEET_BASE_PATH=${MEET_BASE_PATH:-/meet}
+    env_file:
+      - ./meet.env
     volumes:
       - ./custom-layout:/var/www/custom-layout
+      - /etc/localtime:/etc/localtime:ro
     ports:
       - 5443:5443
       - 6443:6443
       - 7443:7443
       - 7880:7880
       - 9443:9443
+      - 9080:9080
     depends_on:
       setup:
         condition: service_completed_successfully
 
   redis:
-    image: docker.io/redis:7.4.4-alpine
-    platform: linux/amd64
+    image: docker.io/redis:8.6.1-alpine
     container_name: redis
     restart: unless-stopped
     ports:
       - 6379:6379
     volumes:
       - redis:/data
+      - /etc/localtime:/etc/localtime:ro
     command: >
-      redis-server
-      --bind 0.0.0.0
-      --requirepass ${REDIS_PASSWORD:-}
+      redis-server --bind 0.0.0.0 --requirepass ${REDIS_PASSWORD:-}
     depends_on:
       setup:
         condition: service_completed_successfully
 
   minio:
-    image: docker.io/openvidu/minio:2025.5.24-debian-12-r1
-    platform: linux/amd64
+    image: docker.io/openvidu/minio:2025.10.15-debian-12-r9
     container_name: minio
     restart: unless-stopped
     ports:
@@ -61,23 +59,25 @@ services:
       - MINIO_ROOT_PASSWORD=${MINIO_SECRET_KEY:-}
       - MINIO_DEFAULT_BUCKETS=openvidu-appdata
       - MINIO_CONSOLE_SUBPATH=/minio-console
+      - MINIO_BROWSER=on
       - MINIO_BROWSER_REDIRECT_URL=http://localhost:7880/minio-console
     volumes:
       - minio-data:/bitnami/minio/data
       - minio-certs:/certs
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   mongo:
-    image: docker.io/openvidu/mongodb:8.0.12-r0
-    platform: linux/amd64
+    image: docker.io/openvidu/mongodb:8.0.19-r1
     container_name: mongo
     restart: unless-stopped
     ports:
       - 27017:27017
     volumes:
       - mongo-data:/bitnami/mongodb
+      - /etc/localtime:/etc/localtime:ro
     environment:
       - MONGODB_ROOT_USER=${MONGO_ADMIN_USERNAME:-}
       - MONGODB_ROOT_PASSWORD=${MONGO_ADMIN_PASSWORD:-}
@@ -85,14 +85,12 @@ services:
       - MONGODB_REPLICA_SET_MODE=primary
       - MONGODB_REPLICA_SET_NAME=rs0
       - MONGODB_REPLICA_SET_KEY=devreplicasetkey
-      - EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU=${EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU:-0}
     depends_on:
       setup:
         condition: service_completed_successfully
 
   dashboard:
-    image: docker.io/openvidu/openvidu-dashboard:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-dashboard:3.6.0
     container_name: dashboard
     restart: unless-stopped
     environment:
@@ -100,13 +98,14 @@ services:
       - ADMIN_USERNAME=${DASHBOARD_ADMIN_USERNAME:-}
       - ADMIN_PASSWORD=${DASHBOARD_ADMIN_PASSWORD:-}
       - DATABASE_URL=mongodb://${MONGO_ADMIN_USERNAME}:${MONGO_ADMIN_PASSWORD}@mongo:27017/?replicaSet=rs0&readPreference=primaryPreferred
+    volumes:
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   openvidu:
-    image: docker.io/openvidu/openvidu-server:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-server:3.6.0
     restart: unless-stopped
     container_name: openvidu
     extra_hosts:
@@ -123,13 +122,13 @@ services:
     volumes:
       - ./livekit.yaml:/etc/livekit.yaml
       - ./scripts/entrypoint_openvidu.sh:/scripts/entrypoint.sh
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   ingress:
-    image: docker.io/openvidu/ingress:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/ingress:3.6.0
     container_name: ingress
     restart: unless-stopped
     extra_hosts:
@@ -142,13 +141,13 @@ services:
       - INGRESS_CONFIG_FILE=/etc/ingress.yaml
     volumes:
       - ./ingress.yaml:/etc/ingress.yaml
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   egress:
-    image: docker.io/livekit/egress:v1.10.0
-    platform: linux/amd64
+    image: docker.io/openvidu/egress:3.6.0
     restart: unless-stopped
     container_name: egress
     extra_hosts:
@@ -158,17 +157,15 @@ services:
     volumes:
       - ./egress.yaml:/etc/egress.yaml
       - egress-data:/home/egress/tmp
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   openvidu-meet:
-    image: docker.io/openvidu/openvidu-meet:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-meet:3.6.0
     container_name: openvidu-meet
     restart: on-failure
-    ports:
-      - 9080:6080
     extra_hosts:
       - host.docker.internal:host-gateway
     environment:
@@ -176,48 +173,34 @@ services:
       - LAN_MODE=${LAN_MODE:-false}
       - LAN_DOMAIN=${LAN_DOMAIN:-}
       - LAN_PRIVATE_IP=${LAN_PRIVATE_IP:-}
-      - SERVER_PORT=6080
-      - MEET_LOG_LEVEL=${MEET_LOG_LEVEL:-info}
-      - MEET_NAME_ID=openviduMeet-LOCAL
-      - MEET_INITIAL_API_KEY=${MEET_INITIAL_API_KEY:-meet-api-key}
-      - MEET_INITIAL_ADMIN_USER=${MEET_INITIAL_ADMIN_USER:-admin}
-      - MEET_INITIAL_ADMIN_PASSWORD=${MEET_INITIAL_ADMIN_PASSWORD:-admin}
-      - MEET_COOKIE_SECURE=false
-      - MEET_INITIAL_WEBHOOK_ENABLED=true
-      - MEET_INITIAL_WEBHOOK_URL=${MEET_INITIAL_WEBHOOK_URL:-http://host.docker.internal:6080/webhook}
-      - LIVEKIT_URL_PRIVATE=ws://openvidu:7880/
       - LIVEKIT_API_KEY=${LIVEKIT_API_KEY}
       - LIVEKIT_API_SECRET=${LIVEKIT_API_SECRET}
-      - MEET_S3_BUCKET=${MEET_S3_BUCKET:-openvidu-appdata}
-      - MEET_S3_SUBBUCKET=${MEET_S3_SUBBUCKET:-openvidu-meet}
-      - MEET_S3_SERVICE_ENDPOINT=${MEET_S3_SERVICE_ENDPOINT:-http://minio:9000}
       - MEET_S3_ACCESS_KEY=${MINIO_ACCESS_KEY}
       - MEET_S3_SECRET_KEY=${MINIO_SECRET_KEY}
-      - MEET_AWS_REGION=${MEET_AWS_REGION:-us-east-1}
-      - MEET_S3_WITH_PATH_STYLE_ACCESS=${MEET_S3_WITH_PATH_STYLE_ACCESS:-true}
-      - MEET_REDIS_HOST=redis
-      - MEET_REDIS_PORT=6379
       - MEET_REDIS_PASSWORD=${REDIS_PASSWORD:-}
-      - MEET_REDIS_DB=0
+      - MEET_MONGO_URI=mongodb://${MONGO_ADMIN_USERNAME}:${MONGO_ADMIN_PASSWORD}@mongo:27017/?replicaSet=rs0&readPreference=primaryPreferred
+      - MEET_BASE_PATH=${MEET_BASE_PATH:-/meet}
+      - MEET_CONFIG_DIR=/config/meet.env
     volumes:
+      - ./meet.env:/config/meet.env
       - ./scripts/entrypoint_openvidu_meet.sh:/scripts/entrypoint.sh
       - ./scripts/utils.sh:/scripts/utils.sh
+      - /etc/localtime:/etc/localtime:ro
     entrypoint: /bin/sh /scripts/entrypoint.sh
     depends_on:
       setup:
         condition: service_completed_successfully
 
   operator:
-    image: docker.io/openvidu/openvidu-operator:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-operator:3.6.0
     container_name: operator
     restart: unless-stopped
     volumes:
       - /var/run/docker.sock:/var/run/docker.sock
       - agents-config:/agents-config
       - ./:/deployment
+      - /etc/localtime:/etc/localtime:ro
     environment:
-      - PLATFORM=linux/amd64
       - MODE=agent-manager-local
       - DEPLOYMENT_FILES_DIR=/deployment
       - AGENTS_CONFIG_DIR=/agents-config
@@ -233,10 +216,11 @@ services:
         condition: service_completed_successfully
 
   ready-check:
-    image: docker.io/openvidu/openvidu-operator:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-operator:3.6.0
     container_name: ready-check
     restart: on-failure
+    volumes:
+      - /etc/localtime:/etc/localtime:ro
     environment:
       - MODE=local-ready-check
       - OPENVIDU_ENVIRONMENT=local-platform
@@ -250,9 +234,8 @@ services:
       - MINIO_SECRET_KEY=${MINIO_SECRET_KEY:-}
       - LIVEKIT_API_KEY=${LIVEKIT_API_KEY:-}
       - LIVEKIT_API_SECRET=${LIVEKIT_API_SECRET:-}
-      - MEET_INITIAL_ADMIN_USER=${MEET_INITIAL_ADMIN_USER:-}
-      - MEET_INITIAL_ADMIN_PASSWORD=${MEET_INITIAL_ADMIN_PASSWORD:-}
-      - MEET_INITIAL_API_KEY=${MEET_INITIAL_API_KEY:-}
+    env_file:
+      - ./meet.env
     depends_on:
       - openvidu
       - ingress
@@ -263,7 +246,6 @@ services:
 
   setup:
     image: docker.io/busybox:1.37.0
-    platform: linux/amd64
     container_name: setup
     restart: "no"
     volumes:
@@ -271,6 +253,7 @@ services:
       - mongo-data:/mongo
       - egress-data:/egress
       - ./scripts/setup.sh:/scripts/setup.sh
+      - /etc/localtime:/etc/localtime:ro
     environment:
       - USE_HTTPS=${USE_HTTPS:-false}
       - LAN_MODE=${LAN_MODE:-false}

community/egress.yaml

@@ -46,8 +46,16 @@ cpu_cost:
 openvidu:
     # Allocation strategy for new egress requests
     # - cpuload: the node with the lowest CPU load will be selected. Distributes the CPU load evenly across all nodes.
-    # - binpack: some node already hosting at least one Egress will be selected. Fills up nodes before assigning work to new ones.
+    # - binpack: some node already hosting at least one egress will be selected. Fills up nodes before assigning work to new ones.
     allocation_strategy: cpuload
+    # Whether to use system-wide CPU monitoring or egress process CPU monitoring. This affects the allocation of new egress requests.
+    # It is preferable to set this value to:
+    # - true: when the egress service is running in a shared server also hosting other CPU-intensive services.
+    # - false: when the egress service is running in a dedicated server.
+    use_global_cpu_monitoring: true
     # Disables the automatic killing of the most expensive egress when CPU is overloaded.
     # The default "false" value helps keeping the node stable, but may cause unexpected egress terminations under high load.
-    disable_cpu_overload_killer: false
+    disable_cpu_overload_killer: false
+    # Minimum available disk space in MB required to accept new egress requests.
+    # Default: 512 MB. Set to a negative value (e.g., -1) to disable disk space checking.
+    min_disk_space_mb: 512

community/meet.env

@@ -0,0 +1,36 @@
+# OpenVidu Meet configuration
+# Static environment variables loaded via MEET_CONFIG_DIR
+
+SERVER_PORT=6080
+MEET_NAME_ID=openviduMeet-LOCAL
+MEET_LOG_LEVEL=info
+MEET_COOKIE_SECURE=false
+MEET_INITIAL_ADMIN_USER=admin
+MEET_INITIAL_ADMIN_PASSWORD=admin
+MEET_INITIAL_API_KEY=meet-api-key
+MEET_INITIAL_WEBHOOK_ENABLED=true
+MEET_INITIAL_WEBHOOK_URL=http://host.docker.internal:6080/webhook
+
+LIVEKIT_URL_PRIVATE=ws://openvidu:7880/
+
+# S3 configuration
+MEET_S3_BUCKET=openvidu-appdata
+MEET_S3_SUBBUCKET=openvidu-meet
+MEET_S3_SERVICE_ENDPOINT=http://minio:9000
+MEET_AWS_REGION=us-east-1
+MEET_S3_WITH_PATH_STYLE_ACCESS=true
+
+# Storage backend type
+MEET_BLOB_STORAGE_MODE=s3
+
+# Redis configuration
+MEET_REDIS_HOST=redis
+MEET_REDIS_PORT=6379
+MEET_REDIS_DB=0
+
+# MongoDB configuration
+MEET_MONGO_ENABLED=true
+MEET_MONGO_DB_NAME=openvidu-meet
+
+# Enable live captions using OpenVidu Speech to Text agent
+MEET_CAPTIONS_ENABLED=false

pro/.env

@@ -30,7 +30,5 @@ MINIO_SECRET_KEY=minioadmin
 MONGO_ADMIN_USERNAME=mongoadmin
 MONGO_ADMIN_PASSWORD=mongoadmin
 
-# OpenVidu Meet configuration.
-MEET_INITIAL_ADMIN_USER=admin
-MEET_INITIAL_ADMIN_PASSWORD=admin
-MEET_INITIAL_API_KEY=meet-api-key
+# OpenVidu Meet base path
+MEET_BASE_PATH=/meet

pro/agent-speech-processing.yaml

@@ -1,5 +1,5 @@
 # Docker image of the agent.
-docker_image: docker.io/openvidu/agent-speech-processing:3.4.0
+docker_image: docker.io/openvidu/agent-speech-processing-vosk:3.6.0
 
 # Whether to run the agent or not.
 enabled: false
@@ -7,18 +7,18 @@ enabled: false
 # Maximum CPU load threshold for the agent to accept new jobs. Value between 0 and 1.
 load_threshold: 1.0
 
-# Log level for the agent [DEBUG, INFO, WARNING, ERROR, CRITICAL]
+# Log level for the agent [DEBUG, INFO, WARN, ERROR, CRITICAL]
 log_level: INFO
 
 live_captions:
-  # How this agent will connect to Rooms [automatic, manual]
-  # - automatic: the agent will automatically connect to new Rooms.
+  # How this agent will connect to Rooms [manual, automatic]
   # - manual: the agent will connect to new Rooms only when your application dictates it by using the Agent Dispatch API.
-  processing: automatic
+  # - automatic: the agent will automatically connect to new Rooms.
+  processing: manual
 
-  # Which speech-to-text AI provider to use [aws, azure, google, openai, groq, deepgram, assemblyai, fal, clova, speechmatics, gladia, sarvam, spitch]
+  # Which speech-to-text AI provider to use [aws, azure, google, openai, azure_openai, groq, deepgram, assemblyai, fal, clova, speechmatics, gladia, sarvam, mistralai, cartesia, soniox, nvidia, elevenlabs, simplismart, vosk, sherpa]
   # The custom configuration for the selected provider must be set below
-  provider:
+  provider: vosk
 
   aws:
     # Credentials for AWS Transcribe. See https://docs.aws.amazon.com/transcribe/latest/dg/what-is.html
@@ -63,6 +63,10 @@ live_captions:
     # Removes profanity (swearing), or replaces letters of profane words with stars. Valid values: Masked | Removed | Raw
     # See https://learn.microsoft.com/en-us/azure/ai-services/translator/profanity-filtering
     profanity:
+    # List of words or phrases to boost recognition accuracy. Azure will give higher priority to these phrases during recognition.
+    phrase_list:
+    # Controls punctuation behavior. If True, enables explicit punctuation mode where punctuation marks are added explicitly. If False (default), uses Azure's default punctuation behavior.
+    explicit_punctuation:
 
   azure_openai:
     # Credentials for Azure OpenAI APIs. See https://learn.microsoft.com/en-us/azure/api-management/api-management-authenticate-authorize-azure-openai
@@ -82,6 +86,8 @@ live_captions:
     project:
     # The language code to use for transcription (e.g., "en" for English).
     language:
+    # Whether to automatically detect the language.
+    detect_language:
     # ID of the model to use for speech-to-text.
     model:
     # Initial prompt to guide the transcription.
@@ -135,6 +141,8 @@ live_captions:
     # The language of the input audio. Supplying the input language in ISO-639-1 format
     # (https://en.wikipedia.org/wiki/List_of_ISO_639-1_codes) will improve accuracy and latency.
     language:
+    # Whether to automatically detect the language.
+    detect_language:
     # Optional text prompt to guide the transcription. Only supported for whisper-1.
     prompt:
 
@@ -146,8 +154,12 @@ live_captions:
     # The language of the input audio. Supplying the input language in ISO-639-1 format
     # (https://en.wikipedia.org/wiki/List_of_ISO_639-1_codes) will improve accuracy and latency.
     language:
+    # Whether to automatically detect the language.
+    detect_language:
     # Prompt to guide the model's style or specify how to spell unfamiliar words. 224 tokens max.
     prompt:
+    # Base URL for the Groq API. By default "https://api.groq.com/openai/v1"
+    base_url:
 
   deepgram:
     # See https://console.deepgram.com/
@@ -156,25 +168,27 @@ live_captions:
     model:
     # See https://developers.deepgram.com/reference/speech-to-text-api/listen-streaming#request.query.language
     language:
-    # Whether to enable automatic language detection. Defaults to false. See https://developers.deepgram.com/docs/language-detection
+    # Whether to enable automatic language detection. See https://developers.deepgram.com/docs/language-detection
     detect_language: false
-    # Whether to return interim (non-final) transcription results. Defaults to true. See https://developers.deepgram.com/docs/interim-results
+    # Whether to return interim (non-final) transcription results. See https://developers.deepgram.com/docs/interim-results
     interim_results: true
-    # Whether to apply smart formatting to numbers, dates, etc. Defaults to false. See https://developers.deepgram.com/docs/smart-format
+    # Whether to apply smart formatting to numbers, dates, etc. See https://developers.deepgram.com/docs/smart-format
     smart_format: false
-    # When smart_format is used, ensures it does not wait for sequence to be complete before returning results. Defaults to true. See https://developers.deepgram.com/docs/smart-format#using-no-delay
+    # When smart_format is used, ensures it does not wait for sequence to be complete before returning results. See https://developers.deepgram.com/docs/smart-format#using-no-delay
     no_delay: true
-    # Whether to add punctuations to the transcription. Defaults to true. Turn detector will work better with punctuations. See https://developers.deepgram.com/docs/punctuation
+    # Whether to add punctuations to the transcription. Turn detector will work better with punctuations. See https://developers.deepgram.com/docs/punctuation
     punctuate: true
-    # Whether to include filler words (um, uh, etc.) in transcription. Defaults to true. See https://developers.deepgram.com/docs/filler-words
+    # Whether to include filler words (um, uh, etc.) in transcription. See https://developers.deepgram.com/docs/filler-words
     filler_words: true
-    # Whether to filter profanity from the transcription. Defaults to false. See https://developers.deepgram.com/docs/profanity-filter
+    # Whether to filter profanity from the transcription. See https://developers.deepgram.com/docs/profanity-filter
     profanity_filter: false
-    # List of tuples containing keywords and their boost values for improved recognition. Each tuple should be (keyword: str, boost: float). Defaults to None. keywords does not work with Nova-3 models. Use keyterms instead.
+    # Whether to transcribe numbers as numerals. See https://developers.deepgram.com/docs/numerals
+    numerals: false
+    # List of tuples containing keywords and their boost values for improved recognition. Each tuple should be (keyword: str, boost: float). keywords does not work with Nova-3 models. Use keyterms instead.
     # keywords:
     #   - [OpenVidu, 1.5]
     #   - [WebRTC, 1]
-    # List of key terms to improve recognition accuracy. Defaults to None. keyterms is supported by Nova-3 models.
+    # List of key terms to improve recognition accuracy. keyterms is supported by Nova-3 models.
     # Commented below is an example
     keyterms:
       # - "OpenVidu"
@@ -183,8 +197,18 @@ live_captions:
   assemblyai:
     # API key for AssemblyAI. See https://www.assemblyai.com/dashboard/api-keys
     api_key:
+    # The confidence threshold (0.0 to 1.0) to use when determining if the end of a turn has been reached.
+    end_of_turn_confidence_threshold:
+    # The minimum amount of silence in milliseconds required to detect end of turn when confident.
+    min_end_of_turn_silence_when_confident:
+    # The maximum amount of silence in milliseconds allowed in a turn before end of turn is triggered.
+    max_turn_silence:
     # Whether to return formatted final transcripts (proper punctuation, letter casing...). If enabled, formatted final transcripts are emitted shortly following an end-of-turn detection.
     format_turns: true
+    # List of keyterms to improve recognition accuracy for specific words and phrases.
+    keyterms_prompt:
+      # - "OpenVidu"
+      # - "WebRTC"
 
   fal:
     # API key for fal. See https://fal.ai/dashboard/keys
@@ -208,12 +232,14 @@ live_captions:
   speechmatics:
     # API key for Speechmatics. See https://portal.speechmatics.com/manage-access/
     api_key:
-    # ISO 639-1 language code. All languages are global and can understand different dialects/accents. To see the list of all supported languages, see https://docs.speechmatics.com/introduction/supported-languages
+    # ISO 639-1 language code. All languages are global and can understand different dialects/accents. To see the list of all supported languages, see https://docs.speechmatics.com/speech-to-text/languages#transcription-languages
     language:
-    # Operating point to use for the transcription per required accuracy & complexity. To learn more, see https://docs.speechmatics.com/features/accuracy-language-packs#accuracy
+    # Operating point to use for the transcription per required accuracy & complexity. To learn more, see https://docs.speechmatics.com/speech-to-text/languages#operating-points
     operating_point:
-    # Partial transcripts allow you to receive preliminary transcriptions and update as more context is available until the higher-accuracy final transcript is returned. Partials are returned faster but without any post-processing such as formatting. See https://docs.speechmatics.com/features/realtime-latency#partial-transcripts
+    # Partial transcripts allow you to receive preliminary transcriptions and update as more context is available until the higher-accuracy final transcript is returned. Partials are returned faster but without any post-processing such as formatting. See https://docs.speechmatics.com/speech-to-text/realtime/output#partial-transcripts
     enable_partials:
+    # Enable speaker diarization. When enabled, the STT engine will determine and attribute words to unique speakers. The speaker_sensitivity parameter can be used to adjust the sensitivity of diarization
+    enable_diarization:
     # RFC-5646 language code to make spelling rules more consistent in the transcription output. See https://docs.speechmatics.com/features/word-tagging#output-locale
     output_locale:
     # The delay in seconds between the end of a spoken word and returning the final transcript results. See https://docs.speechmatics.com/features/realtime-latency#configuration-example
@@ -255,6 +281,10 @@ live_captions:
     languages:
     # Whether to allow switching between languages during recognition. Defaults to True
     code_switching:
+    # https://docs.gladia.io/api-reference/v2/live/init#body-pre-processing-audio-enhancer
+    pre_processing_audio_enhancer:
+    # https://docs.gladia.io/api-reference/v2/live/init#body-pre-processing-speech-threshold
+    pre_processing_speech_threshold:
 
   sarvam:
     # API key for Sarvam. See https://dashboard.sarvam.ai/key-management
@@ -263,3 +293,145 @@ live_captions:
     language:
     # The Sarvam STT model to use. See https://docs.sarvam.ai/api-reference-docs/speech-to-text/transcribe#request.body.model.model
     model:
+
+  mistralai:
+    # API key for Mistral AI. See https://console.mistral.ai/api-keys
+    api_key:
+    # Name of the Voxtral STT model to use. Default to voxtral-mini-latest. See https://docs.mistral.ai/capabilities/audio/
+    model:
+    # The language code to use for transcription (e.g., "en" for English)
+    language:
+
+  cartesia:
+    # API key for Cartesia. See https://play.cartesia.ai/keys
+    api_key:
+    # The Cartesia STT model to use
+    model:
+    # The language code to use for transcription (e.g., "en" for English)
+    language:
+
+  soniox:
+    # API key for Soniox. See https://console.soniox.com/
+    api_key:
+    # Set language hints when possible to significantly improve accuracy. See: https://soniox.com/docs/stt/concepts/language-hints
+    language_hints:
+      # - "en"
+      # - "es"
+    # Set context to improve recognition of difficult and rare words. Context is a string and can include words, phrases, sentences, or summaries (limit: 10K chars). See https://soniox.com/docs/stt/concepts/context
+    context:
+
+  nvidia:
+    # API key for NVIDIA. See https://build.nvidia.com/explore/speech?integrate_nim=true&hosted_api=true&modal=integrate-nim
+    # Required when using NVIDIA's cloud services. To use a self-hosted NVIDIA Riva server setup "server" and "use_ssl" instead.
+    api_key:
+    # The NVIDIA Riva ASR model to use. Default is "parakeet-1.1b-en-US-asr-streaming-silero-vad-sortformer"
+    # See available models: https://build.nvidia.com/search/models?filters=usecase%3Ausecase_speech_to_text
+    model:
+    # The NVIDIA function ID for the model. Default is "1598d209-5e27-4d3c-8079-4751568b1081"
+    function_id:
+    # Whether to add punctuation to transcription results. Default is true.
+    punctuate:
+    # The language code for transcription. Default is "en-US"
+    language_code:
+    # Audio sample rate in Hz. Default is 16000.
+    sample_rate:
+    # The NVIDIA Riva server address. Default is "grpc.nvcf.nvidia.com:443"
+    # For self-hosted NIM, use your server address (e.g., "localhost:50051")
+    server:
+    # Whether to use SSL for the connection. Default is true.
+    # Set to false for locally hosted Riva NIM services without SSL.
+    use_ssl:
+
+  spitch:
+    # API key for Spitch. See https://docs.spitch.app/keys
+    api_key:
+    # Language short code for the generated speech. For supported values, see https://docs.spitch.app/concepts/languages
+    language:
+
+  elevenlabs:
+    # API key for ElevenLabs. See https://elevenlabs.io/app/settings/api-keys
+    api_key:
+    # The ElevenLabs STT model to use. Valid values are ["scribe_v1", "scribe_v2", "scribe_v2_realtime"]. See https://elevenlabs.io/docs/overview/models#models-overview
+    model_id:
+    # An ISO-639-1 or ISO-639-3 language_code corresponding to the language of the audio file. Can sometimes improve transcription performance if known beforehand. Defaults to null, in this case the language is predicted automatically
+    language_code:
+    # Custom base URL for the API. Optional.
+    base_url:
+    # Audio sample rate in Hz. Default is 16000.
+    sample_rate:
+    # Whether to tag audio events like (laughter), (footsteps), etc. in the transcription. Only supported for Scribe v1 model. Default is True
+    tag_audio_events:
+    # Whether to include word-level timestamps in the transcription. Default is false.
+    include_timestamps:
+
+  simplismart:
+    # API key for SimpliSmart. See https://docs.simplismart.ai/model-suite/settings/api-keys
+    api_key:
+    # Model identifier for the backend STT model. One of ["openai/whisper-large-v2", "openai/whisper-large-v3", "openai/whisper-large-v3-turbo"]
+    # Default is "openai/whisper-large-v3-turbo"
+    model:
+    # Language code for transcription (default: "en"). See https://docs.simplismart.ai/get-started/playground/transcription-models#supported-languages-with-their-codes
+    language:
+    # Operation to perform. "transcribe" converts speech to text in the original language, "translate" translates into English. Default is "transcribe".
+    task:
+    # If true, disables timestamp generation in transcripts. Default is true
+    without_timestamps:
+    # Minimum duration (ms) for a valid speech segment. Default is 0
+    min_speech_duration_ms:
+    # Decoding temperature (affects randomness). Default is 0.0
+    temperature:
+    # Whether to permit multilingual recognition. Default is false
+    multilingual:
+
+  vosk:
+    # Vosk language model. This provider requires docker_image "docker.io/openvidu/agent-speech-processing-vosk"
+    # Below is the list of pre-installed models in the container (available at https://alphacephei.com/vosk/models):
+    # - vosk-model-en-us-0.22-lgraph (English US)
+    # - vosk-model-small-cn-0.22 (Chinese)
+    # - vosk-model-small-de-0.15 (German)
+    # - vosk-model-small-en-in-0.4 (English India)
+    # - vosk-model-small-es-0.42 (Spanish)
+    # - vosk-model-small-fr-0.22 (French)
+    # - vosk-model-small-hi-0.22 (Hindi)
+    # - vosk-model-small-it-0.22 (Italian)
+    # - vosk-model-small-ja-0.22 (Japanese)
+    # - vosk-model-small-nl-0.22 (Dutch)
+    # - vosk-model-small-pt-0.3 (Portuguese)
+    # - vosk-model-small-ru-0.22 (Russian)
+    model: vosk-model-en-us-0.22-lgraph
+    # Language code for reference. It has no effect other than observability purposes.
+    # If a pre-installed "model" is declared, this will be set automatically if empty.
+    language:
+    # Audio sample rate in Hz. Default is 16000.
+    sample_rate:
+    # Whether to return interim/partial results during recognition. Default is true.
+    partial_results:
+    # Whether to override Vosk's built-in Voice Activity Detection (VAD) with Silero's VAD. Default is false.
+    use_silero_vad: false
+
+  sherpa:
+    # sherpa streaming model. This provider requires docker_image "docker.io/openvidu/agent-speech-processing-sherpa"
+    # Below is the list of pre-installed models in the container (available at https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models):
+    # - sherpa-onnx-streaming-zipformer-en-kroko-2025-08-06 (English)
+    # - sherpa-onnx-streaming-zipformer-es-kroko-2025-08-06 (Spanish)
+    # - sherpa-onnx-streaming-zipformer-de-kroko-2025-08-06 (German)
+    # - sherpa-onnx-streaming-zipformer-fr-kroko-2025-08-06 (French)
+    # - sherpa-onnx-streaming-zipformer-ar_en_id_ja_ru_th_vi_zh-2025-02-10 (Multilingual: Arabic, English, Indonesian, Japanese, Russian, Thai, Vietnamese, Chinese)
+    model: sherpa-onnx-streaming-zipformer-en-kroko-2025-08-06
+    # Language code for reference. Auto-detected from model name if not set.
+    language:
+    # Runtime provider for sherpa-onnx. Supported values: "cpu" or "cuda". Default is "cpu".
+    # Learn about GPU acceleration at https://openvidu.io/docs/ai/live-captions/#gpu-acceleration-for-sherpa-provider
+    provider:
+    # Audio sample rate in Hz. Default is 16000.
+    sample_rate:
+    # Whether to return interim/partial results during recognition. Default is true.
+    partial_results:
+    # Number of threads for ONNX Runtime. Default is 2.
+    num_threads:
+    # Recognizer type ("transducer", "paraformer", "zipformer_ctc", "nemo_ctc", "t_one_ctc"). Auto-detected from model name if not set.
+    recognizer_type:
+    # Decoding method ("greedy_search", "modified_beam_search"). Default is "greedy_search".
+    decoding_method:
+    # Whether to override sherpa's built-in Voice Activity Detection (VAD) with Silero's VAD. Default is false.
+    use_silero_vad: false

pro/configure_lan_private_ip_macos.sh

@@ -14,14 +14,4 @@ if [ -z "$LAN_PRIVATE_IP" ]; then
 fi
 
 # Replace the LAN_PRIVATE_IP in the .env file
-sed -i'' -e "s/LAN_PRIVATE_IP=.*/LAN_PRIVATE_IP=$LAN_PRIVATE_IP/g" .env
-
-# If sillicon mac, enable EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU flag
-if [ "$(uname -m)" = "arm64" ]; then
-    if ! grep -q "EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU" .env; then
-        echo "# Enable this flag to run Docker Desktop on Apple Silicon Macs" >> .env
-        echo "EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU=1" >> .env
-    else
-        sed -i'' -e "s/EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU=.*/EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU=1/g" .env
-    fi
-fi
+sed -i'' -e "s/LAN_PRIVATE_IP=.*/LAN_PRIVATE_IP=$LAN_PRIVATE_IP/g" .env

pro/docker-compose.yaml

@@ -1,7 +1,6 @@
 services:
   caddy-proxy:
-    image: docker.io/openvidu/openvidu-caddy-local:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-caddy-local:3.6.0
     container_name: caddy-proxy
     restart: unless-stopped
     extra_hosts:
@@ -18,41 +17,40 @@ services:
       - MINIO_ACCESS_KEY=${MINIO_ACCESS_KEY:-}
       - MINIO_SECRET_KEY=${MINIO_SECRET_KEY:-}
       - V2COMPAT_OPENVIDU_SECRET=${LIVEKIT_API_SECRET:-}
-      - MEET_INITIAL_ADMIN_USER=${MEET_INITIAL_ADMIN_USER:-}
-      - MEET_INITIAL_ADMIN_PASSWORD=${MEET_INITIAL_ADMIN_PASSWORD:-}
-      - MEET_INITIAL_API_KEY=${MEET_INITIAL_API_KEY:-}
+      - MEET_BASE_PATH=${MEET_BASE_PATH:-/meet}
+    env_file:
+      - ./meet.env
     volumes:
       - ./custom-layout:/var/www/custom-layout
+      - /etc/localtime:/etc/localtime:ro
     ports:
       - 5443:5443
       - 6443:6443
       - 7443:7443
       - 7880:7880
       - 9443:9443
+      - 9080:9080
     depends_on:
       setup:
         condition: service_completed_successfully
 
   redis:
-    image: docker.io/redis:7.4.4-alpine
-    platform: linux/amd64
+    image: docker.io/redis:8.6.1-alpine
     container_name: redis
     restart: unless-stopped
     ports:
       - 6379:6379
     volumes:
       - redis:/data
+      - /etc/localtime:/etc/localtime:ro
     command: >
-      redis-server
-      --bind 0.0.0.0
-      --requirepass ${REDIS_PASSWORD:-}
+      redis-server --bind 0.0.0.0 --requirepass ${REDIS_PASSWORD:-}
     depends_on:
       setup:
         condition: service_completed_successfully
 
   minio:
-    image: docker.io/openvidu/minio:2025.5.24-debian-12-r1
-    platform: linux/amd64
+    image: docker.io/openvidu/minio:2025.10.15-debian-12-r9
     restart: unless-stopped
     ports:
       - 9000:9000
@@ -61,23 +59,25 @@ services:
       - MINIO_ROOT_PASSWORD=${MINIO_SECRET_KEY:-}
       - MINIO_DEFAULT_BUCKETS=openvidu-appdata
       - MINIO_CONSOLE_SUBPATH=/minio-console
+      - MINIO_BROWSER=on
       - MINIO_BROWSER_REDIRECT_URL=http://localhost:7880/minio-console
     volumes:
       - minio-data:/bitnami/minio/data
       - minio-certs:/certs
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   mongo:
-    image: docker.io/openvidu/mongodb:8.0.12-r0
-    platform: linux/amd64
+    image: docker.io/openvidu/mongodb:8.0.19-r1
     container_name: mongo
     restart: unless-stopped
     ports:
       - 27017:27017
     volumes:
       - mongo-data:/bitnami/mongodb
+      - /etc/localtime:/etc/localtime:ro
     environment:
       - MONGODB_ROOT_USER=${MONGO_ADMIN_USERNAME:-}
       - MONGODB_ROOT_PASSWORD=${MONGO_ADMIN_PASSWORD:-}
@@ -85,14 +85,12 @@ services:
       - MONGODB_REPLICA_SET_MODE=primary
       - MONGODB_REPLICA_SET_NAME=rs0
       - MONGODB_REPLICA_SET_KEY=devreplicasetkey
-      - EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU=${EXPERIMENTAL_DOCKER_DESKTOP_FORCE_QEMU:-0}
     depends_on:
       setup:
         condition: service_completed_successfully
 
   dashboard:
-    image: docker.io/openvidu/openvidu-dashboard:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-dashboard:3.6.0
     container_name: dashboard
     restart: unless-stopped
     environment:
@@ -100,13 +98,14 @@ services:
       - ADMIN_USERNAME=${DASHBOARD_ADMIN_USERNAME:-}
       - ADMIN_PASSWORD=${DASHBOARD_ADMIN_PASSWORD:-}
       - DATABASE_URL=mongodb://${MONGO_ADMIN_USERNAME}:${MONGO_ADMIN_PASSWORD}@mongo:27017/?replicaSet=rs0&readPreference=primaryPreferred
+    volumes:
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   openvidu:
-    image: docker.io/openvidu/openvidu-server-pro:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-server-pro:3.6.0
     restart: unless-stopped
     container_name: openvidu
     extra_hosts:
@@ -125,13 +124,13 @@ services:
     volumes:
       - ./livekit.yaml:/etc/livekit.yaml
       - ./scripts/entrypoint_openvidu.sh:/scripts/entrypoint.sh
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   ingress:
-    image: docker.io/openvidu/ingress:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/ingress:3.6.0
     container_name: ingress
     restart: unless-stopped
     extra_hosts:
@@ -144,13 +143,13 @@ services:
       - INGRESS_CONFIG_FILE=/etc/ingress.yaml
     volumes:
       - ./ingress.yaml:/etc/ingress.yaml
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   egress:
-    image: docker.io/livekit/egress:v1.10.0
-    platform: linux/amd64
+    image: docker.io/openvidu/egress:3.6.0
     restart: unless-stopped
     container_name: egress
     extra_hosts:
@@ -160,17 +159,15 @@ services:
     volumes:
       - ./egress.yaml:/etc/egress.yaml
       - egress-data:/home/egress
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   openvidu-meet:
-    image: docker.io/openvidu/openvidu-meet:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-meet:3.6.0
     container_name: openvidu-meet
     restart: on-failure
-    ports:
-      - 9080:6080
     extra_hosts:
       - host.docker.internal:host-gateway
     environment:
@@ -178,40 +175,26 @@ services:
       - LAN_MODE=${LAN_MODE:-false}
       - LAN_DOMAIN=${LAN_DOMAIN:-}
       - LAN_PRIVATE_IP=${LAN_PRIVATE_IP:-}
-      - SERVER_PORT=6080
-      - MEET_LOG_LEVEL=${MEET_LOG_LEVEL:-info}
-      - MEET_NAME_ID=openviduMeet-LOCAL
-      - MEET_INITIAL_API_KEY=${MEET_INITIAL_API_KEY:-meet-api-key}
-      - MEET_INITIAL_ADMIN_USER=${MEET_INITIAL_ADMIN_USER:-admin}
-      - MEET_INITIAL_ADMIN_PASSWORD=${MEET_INITIAL_ADMIN_PASSWORD:-admin}
-      - MEET_COOKIE_SECURE=false
-      - MEET_INITIAL_WEBHOOK_ENABLED=true
-      - MEET_INITIAL_WEBHOOK_URL=${MEET_INITIAL_WEBHOOK_URL:-http://host.docker.internal:6080/webhook}
-      - LIVEKIT_URL_PRIVATE=ws://openvidu:7880/
       - LIVEKIT_API_KEY=${LIVEKIT_API_KEY}
       - LIVEKIT_API_SECRET=${LIVEKIT_API_SECRET}
-      - MEET_S3_BUCKET=${MEET_S3_BUCKET:-openvidu-appdata}
-      - MEET_S3_SUBBUCKET=${MEET_S3_SUBBUCKET:-openvidu-meet}
-      - MEET_S3_SERVICE_ENDPOINT=${MEET_S3_SERVICE_ENDPOINT:-http://minio:9000}
       - MEET_S3_ACCESS_KEY=${MINIO_ACCESS_KEY}
       - MEET_S3_SECRET_KEY=${MINIO_SECRET_KEY}
-      - MEET_AWS_REGION=${MEET_AWS_REGION:-us-east-1}
-      - MEET_S3_WITH_PATH_STYLE_ACCESS=${MEET_S3_WITH_PATH_STYLE_ACCESS:-true}
-      - MEET_REDIS_HOST=redis
-      - MEET_REDIS_PORT=6379
       - MEET_REDIS_PASSWORD=${REDIS_PASSWORD:-}
-      - MEET_REDIS_DB=0
+      - MEET_MONGO_URI=mongodb://${MONGO_ADMIN_USERNAME}:${MONGO_ADMIN_PASSWORD}@mongo:27017/?replicaSet=rs0&readPreference=primaryPreferred
+      - MEET_BASE_PATH=${MEET_BASE_PATH:-/meet}
+      - MEET_CONFIG_DIR=/config/meet.env
     volumes:
+      - ./meet.env:/config/meet.env
       - ./scripts/entrypoint_openvidu_meet.sh:/scripts/entrypoint.sh
       - ./scripts/utils.sh:/scripts/utils.sh
+      - /etc/localtime:/etc/localtime:ro
     entrypoint: /bin/sh /scripts/entrypoint.sh
     depends_on:
       setup:
         condition: service_completed_successfully
 
   openvidu-v2compatibility:
-    image: docker.io/openvidu/openvidu-v2compatibility:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-v2compatibility:3.6.0
     restart: unless-stopped
     container_name: openvidu-v2compatibility
     entrypoint: /bin/sh /scripts/entrypoint.sh
@@ -247,15 +230,17 @@ services:
       - ./recordings:/opt/openvidu/recordings
       - ./scripts/entrypoint_v2comp.sh:/scripts/entrypoint.sh
       - ./scripts/utils.sh:/scripts/utils.sh
+      - /etc/localtime:/etc/localtime:ro
     depends_on:
       setup:
         condition: service_completed_successfully
 
   ready-check:
-    image: docker.io/openvidu/openvidu-operator:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-operator:3.6.0
     container_name: ready-check
     restart: on-failure
+    volumes:
+      - /etc/localtime:/etc/localtime:ro
     environment:
       - MODE=local-ready-check
       - OPENVIDU_ENVIRONMENT=local-platform
@@ -270,9 +255,8 @@ services:
       - LIVEKIT_API_KEY=${LIVEKIT_API_KEY:-}
       - LIVEKIT_API_SECRET=${LIVEKIT_API_SECRET:-}
       - V2COMPAT_OPENVIDU_SECRET=${LIVEKIT_API_SECRET:-}
-      - MEET_INITIAL_ADMIN_USER=${MEET_INITIAL_ADMIN_USER:-}
-      - MEET_INITIAL_ADMIN_PASSWORD=${MEET_INITIAL_ADMIN_PASSWORD:-}
-      - MEET_INITIAL_API_KEY=${MEET_INITIAL_API_KEY:-}
+    env_file:
+      - ./meet.env
     depends_on:
       - openvidu
       - ingress
@@ -282,16 +266,15 @@ services:
       - mongo
 
   operator:
-    image: docker.io/openvidu/openvidu-operator:3.4.0
-    platform: linux/amd64
+    image: docker.io/openvidu/openvidu-operator:3.6.0
     container_name: operator
     restart: unless-stopped
     volumes:
       - /var/run/docker.sock:/var/run/docker.sock
       - agents-config:/agents-config
       - ./:/deployment
+      - /etc/localtime:/etc/localtime:ro
     environment:
-      - PLATFORM=linux/amd64
       - MODE=agent-manager-local
       - DEPLOYMENT_FILES_DIR=/deployment
       - AGENTS_CONFIG_DIR=/agents-config
@@ -307,7 +290,6 @@ services:
         condition: service_completed_successfully
   setup:
     image: docker.io/busybox:1.37.0
-    platform: linux/amd64
     container_name: setup
     restart: "no"
     volumes:
@@ -315,6 +297,7 @@ services:
       - mongo-data:/mongo
       - egress-data:/egress
       - ./scripts/setup.sh:/scripts/setup.sh
+      - /etc/localtime:/etc/localtime:ro
     environment:
       - USE_HTTPS=${USE_HTTPS:-false}
       - LAN_MODE=${LAN_MODE:-false}

pro/egress.yaml

@@ -46,8 +46,16 @@ cpu_cost:
 openvidu:
     # Allocation strategy for new egress requests
     # - cpuload: the node with the lowest CPU load will be selected. Distributes the CPU load evenly across all nodes.
-    # - binpack: some node already hosting at least one Egress will be selected. Fills up nodes before assigning work to new ones.
+    # - binpack: some node already hosting at least one egress will be selected. Fills up nodes before assigning work to new ones.
     allocation_strategy: cpuload
+    # Whether to use system-wide CPU monitoring or egress process CPU monitoring. This affects the allocation of new egress requests.
+    # It is preferable to set this value to:
+    # - true: when the egress service is running in a shared server also hosting other CPU-intensive services.
+    # - false: when the egress service is running in a dedicated server.
+    use_global_cpu_monitoring: true
     # Disables the automatic killing of the most expensive egress when CPU is overloaded.
     # The default "false" value helps keeping the node stable, but may cause unexpected egress terminations under high load.
     disable_cpu_overload_killer: false
+    # Minimum available disk space in MB required to accept new egress requests.
+    # Default: 512 MB. Set to a negative value (e.g., -1) to disable disk space checking.
+    min_disk_space_mb: 512

pro/meet.env

@@ -0,0 +1,36 @@
+# OpenVidu Meet configuration
+# Static environment variables loaded via MEET_CONFIG_DIR
+
+SERVER_PORT=6080
+MEET_NAME_ID=openviduMeet-LOCAL
+MEET_LOG_LEVEL=info
+MEET_COOKIE_SECURE=false
+MEET_INITIAL_ADMIN_USER=admin
+MEET_INITIAL_ADMIN_PASSWORD=admin
+MEET_INITIAL_API_KEY=meet-api-key
+MEET_INITIAL_WEBHOOK_ENABLED=true
+MEET_INITIAL_WEBHOOK_URL=http://host.docker.internal:6080/webhook
+
+LIVEKIT_URL_PRIVATE=ws://openvidu:7880/
+
+# S3 configuration
+MEET_S3_BUCKET=openvidu-appdata
+MEET_S3_SUBBUCKET=openvidu-meet
+MEET_S3_SERVICE_ENDPOINT=http://minio:9000
+MEET_AWS_REGION=us-east-1
+MEET_S3_WITH_PATH_STYLE_ACCESS=true
+
+# Storage backend type
+MEET_BLOB_STORAGE_MODE=s3
+
+# Redis configuration
+MEET_REDIS_HOST=redis
+MEET_REDIS_PORT=6379
+MEET_REDIS_DB=0
+
+# MongoDB configuration
+MEET_MONGO_ENABLED=true
+MEET_MONGO_DB_NAME=openvidu-meet
+
+# Enable live captions using OpenVidu Speech to Text agent
+MEET_CAPTIONS_ENABLED=false