Bump to version 3.6.0

Add sherpa.provider YAML property to agent-speech-processing.yaml
Minor update of YAML comment
2026-03-06 12:05:23 +00:00 · 2026-03-04 13:09:53 +01:00 · 2026-02-25 18:30:18 +01:00 · 2026-02-25 18:22:19 +01:00 · 2026-02-24 15:29:10 +01:00 · 2026-02-20 14:10:57 +01:00
10 changed files with 388 additions and 110 deletions
--- a/community/.env
+++ b/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
--- a/community/agent-speech-processing.yaml
+++ b/community/agent-speech-processing.yaml
@ -1,5 +1,5 @@
 # Docker image of the agent.
-docker_image: docker.io/openvidu/agent-speech-processing:3.5.0
+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, azure_openai, groq, deepgram, assemblyai, fal, clova, speechmatics, gladia, sarvam, mistralai, cartesia, soniox]
+  # 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
@ -66,7 +66,7 @@ live_captions:
    # 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: 
+    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
@ -318,4 +318,120 @@ live_captions:
      # - "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:
+    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
--- a/community/docker-compose.yaml
+++ b/community/docker-compose.yaml
@ -1,6 +1,6 @@
 services:
  caddy-proxy:
-    image: docker.io/openvidu/openvidu-caddy-local:3.5.0
+    image: docker.io/openvidu/openvidu-caddy-local:3.6.0
    container_name: caddy-proxy
    restart: unless-stopped
    extra_hosts:
@ -16,9 +16,9 @@ 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
@ -28,12 +28,13 @@ services:
      - 7443:7443
      - 7880:7880
      - 9443:9443
+      - 9080:9080
    depends_on:
      setup:
        condition: service_completed_successfully

  redis:
-    image: docker.io/redis:8.2.2-alpine
+    image: docker.io/redis:8.6.1-alpine
    container_name: redis
    restart: unless-stopped
    ports:
@ -42,15 +43,13 @@ services:
      - 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:2025.9.7-debian-12-r3
+    image: docker.io/openvidu/minio:2025.10.15-debian-12-r9
    container_name: minio
    restart: unless-stopped
    ports:
@ -71,7 +70,7 @@ services:
        condition: service_completed_successfully

  mongo:
-    image: docker.io/openvidu/mongodb:8.0.15-r0
+    image: docker.io/openvidu/mongodb:8.0.19-r1
    container_name: mongo
    restart: unless-stopped
    ports:
@ -91,7 +90,7 @@ services:
        condition: service_completed_successfully

  dashboard:
-    image: docker.io/openvidu/openvidu-dashboard:3.5.0
+    image: docker.io/openvidu/openvidu-dashboard:3.6.0
    container_name: dashboard
    restart: unless-stopped
    environment:
@ -106,7 +105,7 @@ services:
        condition: service_completed_successfully

  openvidu:
-    image: docker.io/openvidu/openvidu-server:3.5.0
+    image: docker.io/openvidu/openvidu-server:3.6.0
    restart: unless-stopped
    container_name: openvidu
    extra_hosts:
@ -129,7 +128,7 @@ services:
        condition: service_completed_successfully

  ingress:
-    image: docker.io/openvidu/ingress:3.5.0
+    image: docker.io/openvidu/ingress:3.6.0
    container_name: ingress
    restart: unless-stopped
    extra_hosts:
@ -148,7 +147,7 @@ services:
        condition: service_completed_successfully

  egress:
-    image: docker.io/openvidu/egress:3.5.0
+    image: docker.io/openvidu/egress:3.6.0
    restart: unless-stopped
    container_name: egress
    extra_hosts:
@ -164,11 +163,9 @@ services:
        condition: service_completed_successfully

  openvidu-meet:
-    image: docker.io/openvidu/openvidu-meet:3.5.0
+    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:
@ -176,31 +173,16 @@ 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
@ -210,7 +192,7 @@ services:
        condition: service_completed_successfully

  operator:
-    image: docker.io/openvidu/openvidu-operator:3.5.0
+    image: docker.io/openvidu/openvidu-operator:3.6.0
    container_name: operator
    restart: unless-stopped
    volumes:
@ -234,7 +216,7 @@ services:
        condition: service_completed_successfully

  ready-check:
-    image: docker.io/openvidu/openvidu-operator:3.5.0
+    image: docker.io/openvidu/openvidu-operator:3.6.0
    container_name: ready-check
    restart: on-failure
    volumes:
@ -252,9 +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
--- a/community/egress.yaml
+++ b/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
--- a/community/meet.env
+++ b/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
--- a/pro/.env
+++ b/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
--- a/pro/agent-speech-processing.yaml
+++ b/pro/agent-speech-processing.yaml
@ -1,5 +1,5 @@
 # Docker image of the agent.
-docker_image: docker.io/openvidu/agent-speech-processing:3.5.0
+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, azure_openai, groq, deepgram, assemblyai, fal, clova, speechmatics, gladia, sarvam, mistralai, cartesia, soniox]
+  # 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
@ -66,7 +66,7 @@ live_captions:
    # 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: 
+    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
@ -318,4 +318,120 @@ live_captions:
      # - "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:
+    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
--- a/pro/docker-compose.yaml
+++ b/pro/docker-compose.yaml
@ -1,6 +1,6 @@
 services:
  caddy-proxy:
-    image: docker.io/openvidu/openvidu-caddy-local:3.5.0
+    image: docker.io/openvidu/openvidu-caddy-local:3.6.0
    container_name: caddy-proxy
    restart: unless-stopped
    extra_hosts:
@ -17,9 +17,9 @@ 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
@ -29,12 +29,13 @@ services:
      - 7443:7443
      - 7880:7880
      - 9443:9443
+      - 9080:9080
    depends_on:
      setup:
        condition: service_completed_successfully

  redis:
-    image: docker.io/redis:8.2.2-alpine
+    image: docker.io/redis:8.6.1-alpine
    container_name: redis
    restart: unless-stopped
    ports:
@ -43,15 +44,13 @@ services:
      - 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:2025.9.7-debian-12-r3
+    image: docker.io/openvidu/minio:2025.10.15-debian-12-r9
    restart: unless-stopped
    ports:
      - 9000:9000
@ -71,7 +70,7 @@ services:
        condition: service_completed_successfully

  mongo:
-    image: docker.io/openvidu/mongodb:8.0.15-r0
+    image: docker.io/openvidu/mongodb:8.0.19-r1
    container_name: mongo
    restart: unless-stopped
    ports:
@ -91,7 +90,7 @@ services:
        condition: service_completed_successfully

  dashboard:
-    image: docker.io/openvidu/openvidu-dashboard:3.5.0
+    image: docker.io/openvidu/openvidu-dashboard:3.6.0
    container_name: dashboard
    restart: unless-stopped
    environment:
@ -106,7 +105,7 @@ services:
        condition: service_completed_successfully

  openvidu:
-    image: docker.io/openvidu/openvidu-server-pro:3.5.0
+    image: docker.io/openvidu/openvidu-server-pro:3.6.0
    restart: unless-stopped
    container_name: openvidu
    extra_hosts:
@ -131,7 +130,7 @@ services:
        condition: service_completed_successfully

  ingress:
-    image: docker.io/openvidu/ingress:3.5.0
+    image: docker.io/openvidu/ingress:3.6.0
    container_name: ingress
    restart: unless-stopped
    extra_hosts:
@ -150,7 +149,7 @@ services:
        condition: service_completed_successfully

  egress:
-    image: docker.io/openvidu/egress:3.5.0
+    image: docker.io/openvidu/egress:3.6.0
    restart: unless-stopped
    container_name: egress
    extra_hosts:
@ -166,11 +165,9 @@ services:
        condition: service_completed_successfully

  openvidu-meet:
-    image: docker.io/openvidu/openvidu-meet:3.5.0
+    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:
@ -178,31 +175,16 @@ 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
@ -212,7 +194,7 @@ services:
        condition: service_completed_successfully

  openvidu-v2compatibility:
-    image: docker.io/openvidu/openvidu-v2compatibility:3.5.0
+    image: docker.io/openvidu/openvidu-v2compatibility:3.6.0
    restart: unless-stopped
    container_name: openvidu-v2compatibility
    entrypoint: /bin/sh /scripts/entrypoint.sh
@ -254,7 +236,7 @@ services:
        condition: service_completed_successfully

  ready-check:
-    image: docker.io/openvidu/openvidu-operator:3.5.0
+    image: docker.io/openvidu/openvidu-operator:3.6.0
    container_name: ready-check
    restart: on-failure
    volumes:
@ -273,9 +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
@ -285,7 +266,7 @@ services:
      - mongo

  operator:
-    image: docker.io/openvidu/openvidu-operator:3.5.0
+    image: docker.io/openvidu/openvidu-operator:3.6.0
    container_name: operator
    restart: unless-stopped
    volumes:
--- a/pro/egress.yaml
+++ b/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
--- a/pro/meet.env
+++ b/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
Author	SHA1	Message	Date
GitHub Actions	3425341e81	Bump to version 3.6.0	2026-03-06 12:05:23 +00:00
pabloFuente	8a4923e4a8	Add sherpa.provider YAML property to agent-speech-processing.yaml	2026-03-04 13:09:53 +01:00
pabloFuente	ea27ec1f96	Minor update of YAML comment	2026-02-25 18:30:18 +01:00
cruizba	c145494bbd	Change speech processing default from automatic to manual The agent will now only connect to Rooms when explicitly requested via the Agent Dispatch API instead of auto-connecting to all new Rooms.	2026-02-25 18:22:19 +01:00
cruizba	9bff284b8f	Bump Docker image versions for Redis, Minio and MongoDB - Redis: 8.2.2-alpine -> 8.6.1-alpine - Minio: 2025.9.7-debian-12-r3 -> 2025.10.15-debian-12-r9 - MongoDB: 8.0.15-r0 -> 8.0.19-r1	2026-02-24 15:29:10 +01:00
pabloFuente	1724fa5c18	Add use_global_cpu_monitoring and min_disk_space_mb to egress config	2026-02-20 14:10:57 +01:00
pabloFuente	3d06d98ea0	Beautify docker-compose.yaml	2026-02-18 13:19:34 +01:00
pabloFuente	ff54026aad	Default live_captions provider to vosk	2026-02-12 12:33:31 +01:00
cruizba	c778720ba5	Add meet.env file to caddy-proxy and ready-check which needs `MEET_INITIAL_*` env variables.	2026-02-09 20:03:23 +01:00
cruizba	0422cbd8c2	Refactor OpenVidu Meet configuration: consolidate environment variables into meet.env and update docker-compose to use new configuration structure	2026-02-09 19:49:59 +01:00
pabloFuente	7c22e68ab5	Update agent-speech-processing.yaml files	2026-02-06 10:57:12 +01:00
cruizba	1fd49f308c	Add MEET_BASE_PATH environment variable to configuration files	2026-02-05 19:12:34 +01:00
pabloFuente	8fc6edaa87	Fix agent-speech-processing.yaml files sherpa model names	2026-02-04 17:04:24 +01:00
pabloFuente	803dfbbfa8	Updated agent-speech-processing.yaml files with new providers	2026-02-04 16:39:15 +01:00
cruizba	0ee45ec06f	Route 9080 through caddy to redir to `/meet`	2026-02-02 22:00:22 +01:00
pabloFuente	2ea399dc42	Add nvidia and vosk live_captions providers to YAML	2026-01-21 11:48:59 +01:00
pabloFuente	d51a1b2cdf	Update agent-speech-processing.yaml	2026-01-19 14:07:35 +01:00
cruizba	7c8908707b	Revert to main version	2026-01-10 03:08:48 +01:00