APM — Distributed Tracing

• Area: Observability
• Path: observe/apm (engine) + platform/apm (SaaS)
• Kind: Distributed tracing / Application Performance Monitoring

Role in the stack

APM is the distributed tracing and performance monitoring layer of Koder. Services emit spans via OTLP, Zipkin, Jaeger or simple JSON; APM stitches them into traces, visualizes them as waterfalls, builds service dependency maps, and computes P50P95P99 latency per service. It is the go-to tool when latency spikes or an error flood appears.

The open-core engine lives in observe/apm; platform/apm is the multi-tenant SaaS version.

Features

• Multi-format ingestion — OTLP (OpenTelemetry), Zipkin v2, Jaeger, and simple JSON
• Waterfall visualization — Gantt-chart view of all spans in a trace with timing
• Service dependency map — Visual graph of service-to-service calls with error rates
• Latency analytics — P50P95P99 percentiles, throughput, error rates per service
• Multi-tenant — Org-based isolation with API key authentication
• SQLite storage — Zero-dependency persistence with retention cleanup

Primary couplings