Skip to main content

Voice agent workers (livekit-agents)

LiveKit Agents workers that answer the calls routed to a Navin voice agent: registered under the agent name of the SIP dispatch rules, they fetch the configuration of the agent from Navin Enterprise, talk through OpenAI compatible speech to text, LLM (LiteLLM) and text to speech endpoints, transfer to a human and report the call events.

Generated from catalog/services/livekit-agents.yaml: do not edit by hand.

Summary​

Version1.0.0
Useinstall and configure
Layerfunctional (AI catalog)
CategoryAI serving
Installation methodspython
External connectionno
Sandbox onlyno

Licence​

SPDX identifierApache-2.0
Licence policycore licence (MIT or Apache-2.0)
Licence exception-
Licence exceptions of its dependenciesLEX-029 (pending, awaiting the client), LEX-030 (pending, awaiting the client), LEX-031 (pending, awaiting the client)
Licence noteThe LiveKit Agents SDK and its OpenAI and Silero plugins are Apache-2.0 (https://github.com/livekit/agents); the navin-voice-agent package is the code of Navin.
Platform componentlivekit-agents

Packages and artifacts​

A package the project of the service publishes (its image, its binary archive, its Python or OS package) carries the licence of the service; a third-party package carries the licence written on its row, read from the official upstream metadata.

KindPackageVersionLicenceDetails
Python packagenavin-voice-agent1.0.0LicenseRef-Navin-Enterpriseinterpreters, in order of preference python3.12, python3.11, python3.10
Python packagelivekit-agents1.8.10Apache-2.0interpreters, in order of preference python3.12, python3.11, python3.10
Python packagelivekit-plugins-openai1.8.10Apache-2.0interpreters, in order of preference python3.12, python3.11, python3.10
Python packagelivekit-plugins-silero1.8.10Apache-2.0interpreters, in order of preference python3.12, python3.11, python3.10

Supported systems and architectures​

MethodOperating systemsArchitectures
pythonRHEL family (dnf), Debian family (apt), Arch Linux (pacman)x86_64, aarch64 (not restricted by the descriptor)

Certified host systems: RHEL 8/9/10, Rocky Linux 8/9/10, AlmaLinux 8/9/10, Oracle Linux 8/9, CentOS Stream 9/10, Ubuntu 22.04/24.04, Debian 12, Arch Linux (rolling). The host preflight accepts these systems (other releases of a known family give a warning, end of life releases are refused) and the x86_64 and aarch64 architectures.

Prerequisites​

Host​

  • at least 2048 MB of memory (warning, memory).
  • at least 2 CPU (warning, cpu).

Network and ports​

ComponentPort namePortOpened in the host firewall
workerhealth8581/tcpno
workermetrics6793/tcpyes

Services and connections needed​

Details and resolution: see Dependencies.

Installation methods and modes​

MethodDefaultWhat the method uses
pythonyesa Python virtual environment holding the Python packages of the descriptor

Components​

ComponentKindInstallation methodsPorts
Voice agent worker (worker)workerpython8581/tcp, 6793/tcp

Single worker (single) (default)​

One worker host.

What the mode deploys:

ComponentPlacement
workermin 1, max 1

Worker pool (pool)​

Several workers registered under the same agent name; LiveKit gives each new call to a worker below its load threshold.

What the mode deploys:

ComponentPlacement
workermin 2, recommended 2

Install this service​

The install wizard of the Navin Manager plans, validates and runs the installation; you need the services:install permission on the cluster.

  1. Open the wizard. In Clusters, choose the row action Install services of the target cluster. The wizard opens on its first step, Service.

  2. Choose the service. On the Service step, select Voice agent workers. The card shows the licence Apache-2.0 and the dependencies the installation needs.

  3. Choose the hosts and the mode. On the Hosts step, tick the hosts, then choose the Mode: single (default), pool. Each mode deploys the components listed in Installation methods and modes.

  4. Bind the dependencies. On the Bindings step, choose for each dependency (livekit, navin_enterprise, litellm) Install with the platform or Use an existing connection; an optional dependency can stay on No binding.

  5. Configure. On the Configure step, review the parameters (see Configuration parameters). Passwords left empty with a generated default are generated and sealed by the Manager.

  6. Review and install. On the Install step the Manager validates the plan and shows the installation order; when it says "The plan is valid. Ready to install.", choose Install. A highly available mode is installed only when its HA checks pass.

  7. Verify. The operation ends when the health checks of every instance pass (see Health checks); the service then shows as running in Services.

High availability​

Clusternative (native)
High availabilityredundancy (redundancy)
Failovernative (native)
Replicationnone (none)
Quorumnone (none)
Discoveryregistry (registry)
Entry pointnone (none)
Official documentationhttps://docs.livekit.io/agents/server/
Referenceshttps://github.com/livekit/agents

Topology per mode​

ModesHighest catalog levelDistributionQuorum componentFault domainsDescription
singlestandalone (standalone)single (single)-1One worker, no fault tolerance.
poolha (ha)parallel jobs (parallel-jobs)-2Several workers; LiveKit dispatches each new call to an available worker.

Notes​

  • The loss of a worker ends the voice sessions it runs; the fallback rule of Navin Enterprise routes those calls to a human queue.

Configuration parameters​

Network​

ParameterTypeDefaultAllowed valuesRequiredSecretRestartDescription
Health port (health_port)port8581-nononoHTTP health of the worker: 200 once registered with LiveKit, 503 otherwise; /worker gives its agent name and load.
Metrics port (metrics_port)port6793-nononoPrometheus port of the worker.
LiveKit URL (livekit_url)string(empty)-nononows:// or wss:// URL of LiveKit Server (its load balancer in the distributed mode); empty: the first server of the livekit binding.

Models​

ParameterTypeDefaultAllowed valuesRequiredSecretRestartDescription
Agent name (agent_name)stringnavin-voice^[a-z][a-z0-9-]{1,62}$nonoyesName the dispatch rules of Navin Enterprise give in room_config.agents (explicit dispatch).
LLM key (llm_api_key)password(empty)-noyesnoLiteLLM virtual key of the workers (sk-...); empty: the master key of the litellm binding.
Speech key (speech_api_key)password(empty)-noyesnoKey of the speech to text and text to speech endpoints; empty: the LLM key.

Capacity​

ParameterTypeDefaultAllowed valuesRequiredSecretRestartDescription
Load threshold (load_threshold)float0.7min 0.1, max 1.0nononoCPU load above which a worker takes no new call.
Warm processes (idle_processes)int2min 0, max 64nononoJob processes kept warm (VAD loaded) to answer without delay.

Dependencies​

For each active dependency the Manager takes the first of: the binding chosen in the wizard; a service of the same installation that provides it; a service already installed in the cluster; the only connection of the expected type. Otherwise an optional dependency stays unbound, and a required one is added to the installation in its default mode (the column below names the catalog service added).

BindingProvided byConnection typeEndpointOptionalWhen nothing is boundDescription
livekitlivekit-signalnolivekit is added to the installationLiveKit Server the workers register with (API key pair).
navin_enterprisenavin-enterprise-apinonavin-enterprise is added to the installationTelephony API of Navin Enterprise: runtime of the voice agents, call events, transfers and lookups (worker token of the binding).
litellmlitellm-apinolitellm is added to the installationOpenAI compatible gateway of the LLM and, when Navin Enterprise routes them there, of the speech models.

Commands​

CommandComponentTargetDescription
Worker state (worker_info)workerevery hostAgent name, active calls and load of every worker.
Recent logs (logs)workerchosen hostLast lines of the worker journal.

Health checks​

Liveness (per instance)​

ComponentCheckDetails
workerHTTP http://{{ host.ip or '127.0.0.1' }}:{{ config.health_port }}/Expected 200; 60 retries; interval 2 s; timeout 3 s

Preflight checks​

CheckTypeMeaningSeverityComponents
memorymemory_mb {"min": 2048}at least 2048 MB of memorywarningevery component
cpucpu_count {"min": 2}at least 2 CPUwarningevery component

Uninstall​

Paths removed when the data is purged: {{ paths.log }}.

  • Metrics: http://{{ host.fqdn }}:{{ config.metrics_port }}/metrics

Upstream documentation​