Quickstart

What you’ll get

ObservabilityProvider auto-instruments LLMClientProtocol and VectorStoreProtocol with distributed tracing, metrics, and health checks — zero code changes to your AI logic.

Install

uv add lexigram-ai-observability

Optional OpenTelemetry export:

uv add "lexigram-ai-observability[opentelemetry]"

Minimal Example

from lexigram import Application, LexigramConfig
from lexigram.ai.observability import ObservabilityModule

async def main() -> None:
    app = Application(name="my-ai-app", config=LexigramConfig.from_yaml())

    app.add_module(ObservabilityModule.configure())
    await app.start()

    # Everything registered — LLM/vector calls are now traced
    await app.stop()

What Just Happened

1. ObservabilityProvider registered AITracer, AIMetrics, and AIHealthMonitor as container singletons.
2. During boot(), the provider detected any registered LLMClientProtocol and VectorStoreProtocol instances and wrapped them with ObservableLLMClient / ObservableVectorStore.
3. Subsequent LLM completions and vector store operations are automatically traced and metered.

Wiring with Other AI Packages

from lexigram import Application, LexigramConfig
from lexigram.ai.observability import ObservabilityModule
from lexigram.ai.llm import LLMModule

config = LexigramConfig.from_yaml()
app = Application(name="observable-ai", config=config)
app.add_module(LLMModule.configure())
app.add_module(ObservabilityModule.configure())
await app.start()

Next Steps

• Guide — tracing, metrics, and health checks in depth
• How-Tos — practical recipes
• Configuration — every config key