Configuration

Config Key

The configuration section is ai_observability (loaded from the ai_observability: key in application.yaml).

Options

Key	Type	Default	Env Variable	Description
enabled	bool	true	LEX_AI_OBSERVABILITY__ENABLED	Master on/off switch for all AI observability
metrics_enabled	bool	true	LEX_AI_OBSERVABILITY__METRICS_ENABLED	Enable metrics collection (counters, histograms, gauges)
tracing_enabled	bool	true	LEX_AI_OBSERVABILITY__TRACING_ENABLED	Enable distributed tracing (span creation and export)
health_checks_enabled	bool	true	LEX_AI_OBSERVABILITY__HEALTH_CHECKS_ENABLED	Enable background health checking for AI components

Example YAML

ai_observability:
  enabled: true
  metrics_enabled: true
  tracing_enabled: true
  health_checks_enabled: true

Minimal (production-hardened) YAML

ai_observability:
  enabled: true
  metrics_enabled: true
  tracing_enabled: true
  health_checks_enabled: true

Env Variable Override

export LEX_AI_OBSERVABILITY__ENABLED=true
export LEX_AI_OBSERVABILITY__TRACING_ENABLED=true
export LEX_AI_OBSERVABILITY__METRICS_ENABLED=false

Production Warnings

When Environment.PRODUCTION is detected, the config’s validate_for_environment() method emits warnings if tracing_enabled or metrics_enabled are false — these features are expected in production for operational visibility.

Programmatic

from lexigram.ai.observability.config import ObservabilityConfig

config = ObservabilityConfig(
    enabled=True,
    metrics_enabled=True,
    tracing_enabled=False,
    health_checks_enabled=True,
)

Config Model

Loaded as a BaseConfig subclass (ObservabilityConfig) by ObservabilityProvider. The config instance is registered as a container singleton at register() time and can be resolved by other services:

cfg = await container.resolve(ObservabilityConfig)