OpenTelemetry Backends

PURISTA emits OpenTelemetry traces, metrics, and logs automatically for every command, subscription, stream, and queue job. You choose where that data goes by wiring a SpanProcessor at startup — no changes to business logic, no instrumentation required.

How it works in PURISTA

PURISTA instruments traces by accepting a SpanProcessor at construction time. You create a SpanProcessor wrapping your chosen exporter and pass it directly to the event bridge and each service — no global tracer registration required:

import { OTLPTraceExporter } from '@opentelemetry/exporter-trace-otlp-http'
import { SimpleSpanProcessor } from '@opentelemetry/sdk-trace-node'
import { AmqpBridge } from '@purista/amqpbridge'

const spanProcessor = new SimpleSpanProcessor(
  new OTLPTraceExporter({ url: 'http://localhost:4318/v1/traces' })
)

const eventBridge = new AmqpBridge({ spanProcessor })
const myService = await myV1Service.getInstance(eventBridge, { spanProcessor })

If you also want auto-instrumentation for Node.js built-ins (HTTP, fetch, database drivers), initialize NodeSDK before starting PURISTA:

import { NodeSDK } from '@opentelemetry/sdk-node'
import { OTLPTraceExporter } from '@opentelemetry/exporter-trace-otlp-http'

const sdk = new NodeSDK({
  traceExporter: new OTLPTraceExporter({ url: 'http://localhost:4318/v1/traces' }),
})
sdk.start()
// Then start your PURISTA event bridge and services

Both approaches are compatible — use NodeSDK for broad auto-instrumentation and the SpanProcessor injection for PURISTA-specific span control.

Supported backends

Choose the backend that fits your stack. Every one uses the same SpanProcessor pattern above — only the exporter changes.

For PURISTA AI Harness systems, CloudGrid AI Evaluation is the product layer around the harness eval helpers: datasets, experiment runs, per-item score records, comparisons, optimization candidates, and trace-backed evidence live beside the telemetry they depend on.

Choosing a backend

flowchart TD
    Q["Where do you run?"] --> A["Local dev / open source"]
    Q --> B["Already using Grafana"]
    Q --> C["AWS infrastructure"]
    Q --> D["Azure infrastructure"]
    Q --> E["GCP infrastructure"]
    A --> A1["Jaeger or SigNoz<br/>— single Docker container"]
    B --> B1["Grafana Tempo<br/>— Loki + trace correlation"]
    C --> C1["AWS X-Ray<br/>— via ADOT Collector"]
    D --> D1["Azure Monitor<br/>— Application Insights"]
    E --> E1["Google Cloud Trace<br/>— native GCP exporter"]

Graceful shutdown

Always flush spans before your process exits:

gracefulShutdown(logger, [
  eventBridge,
  myService,
  {
    name: 'OTSpanProcessor',
    destroy: () => spanProcessor.shutdown(),
  },
])

Next steps

Pick a backend from the list in the left sidebar, or start with Jaeger for the fastest local setup.