Wire Formats & Receivers

Most teams adopting InfraSage already have telemetry flowing somewhere. The fastest onboarding is usually not "re-instrument everything." It's pointing what you already run at an endpoint here.

OpenTelemetry (OTLP/HTTP)

The primary path. Point any OpenTelemetry SDK or Collector at:

POST https://api.infrasage.dev/v1/traces
POST https://api.infrasage.dev/v1/metrics
POST https://api.infrasage.dev/v1/logs

Encodings accepted (the stock Collector defaults work as-is):

Aspect	Accepted
Content-Type	application/x-protobuf (SDK/Collector default) or application/json
Content-Encoding	gzip (Collector default), deflate, or none

A minimal Collector exporter needs no special encoding settings:

exporters:
  otlphttp:
    endpoint: https://api.infrasage.dev
    headers:
      X-API-Key: "${INFRASAGE_API_KEY}"

:::note OTLP/gRPC The gRPC transport (:4317) is not yet accepted. Collectors and SDKs configured for gRPC should use the otlphttp exporter instead. Same data, same fidelity. :::

What happens to your metrics on arrival

Two conversions are applied at the boundary, and both matter for correctness (see metric fidelity): cumulative counters become per-interval deltas (with counter-reset detection), and histogram buckets become p50/p95/p99 series alongside count and average.

Already running Prometheus?

Add a remote_write block, with no collector in between:

remote_write:
  - url: https://api.infrasage.dev/api/v1/receiver/prometheus
    headers:
      X-API-Key: "<your key>"

Your exporters (node_exporter, kube-state-metrics, and everything else already scraped) start flowing without re-instrumentation.

Other receivers

Source	Endpoint / mechanism
Splunk HEC	POST /api/v1/receiver/splunk (set SPLUNK_HEC_TOKEN)
Elasticsearch bulk	POST /api/v1/receiver/elasticsearch/_bulk
Datadog agent	Datadog receiver
Kafka	Consumer on your topic
AWS Kinesis Firehose	Firehose receiver
AWS CloudWatch	Metric + log polling
Azure / GCP	Cloud monitor polling
Kubernetes events	Cluster event receiver
SNMP	Interface counters and error rates from network gear
IPMI	BMC sensors: temperature, fans, power draw
CI systems	Deploy and build events
Alertmanager	Webhook (keeps your existing rules working during migration)

Infrastructure metrics

If your Collector isn't already collecting host metrics, enable the hostmetrics receiver. It is what feeds the saturation detector:

receivers:
  hostmetrics:
    collection_interval: 30s
    scrapers:
      cpu:
        metrics:
          system.cpu.utilization: { enabled: true }
      memory:
        metrics:
          system.memory.utilization: { enabled: true }
      filesystem:
        metrics:
          system.filesystem.utilization: { enabled: true }

:::tip Per-instance identity InfraSage derives instance identity from k8s.pod.name, k8s.pod.uid, faas.instance, container.id, host.name, or host.id, in that order. It does not read service.instance.id. If your host metrics arrive without an instance, add one of the supported attributes via a resource processor; without it, a full disk on one node averages away across the fleet. :::