Skip to main content
Terms as they are used here. Where a word means something specific in aiAxonIQ, that meaning is the one below.

​
The pairs people confuse

Worth reading even if you skip the rest.

​
A–Z

The eBPF process that observes services from the kernel and exports OTLP. Upstream OpenTelemetry eBPF Instrumentation, unmodified. See Zero-code instrumentation.
Kernel-level observation of what AI agents do — processes spawned, connections opened, files touched. A different product from the zero-code agent. See AgentSight.
A key-value pair on telemetry. Resource attributes describe the emitter and appear on every record it produces; span and log attributes describe one event. See Resource attributes.
A role that reads the audit log and can change nothing. Deliberately not a rung on the Viewer → Editor → Admin → Owner ladder. See Roles.
How fast an error budget is being consumed, relative to the rate that would exactly exhaust it over the window. See SLOs.
A named Linux capability set for the zero-code agent — A, B or C. The privilege decision, made explicitly. See Capability profiles.
The number of distinct label combinations on a metric. Every combination is a separate stored series, which is what drives metric cost — not the number of samples. See cardinality.
The OpenTelemetry Collector: receives, transforms and forwards telemetry, and collects host and container metrics itself. See Collector configuration.
A Linux facility for running verified programs inside the kernel, attached to events. What makes zero-code instrumentation possible. See How eBPF works.
The amount of failure an SLO permits — the gap between the target and 100%. See SLOs.
The credential authenticating ingest, sent as X-License-Key. Decides which organization data lands in. Shown once. See License keys.
Streaming telemetry as it arrives. Not a search over history — nothing that arrived before you connected appears. See Live tail.
The template behind many similar log lines, discovered by clustering bodies. See Log patterns.
The unit everything belongs to — telemetry, keys, users, plan. The isolation boundary. See Organizations.
OpenTelemetry Protocol, the wire format for logs, metrics and traces. Over HTTP or gRPC. See Ingest endpoints.
Rate, Errors, Duration — the three numbers that describe a service’s health. Derived from spans, so instrumenting for traces produces them for free.
A pre-aggregated metric table at one-minute or one-hour resolution. Why a wide time range answers at coarser resolution, and why metrics outlive logs. See Data retention.
Keeping a fraction of traces. Head sampling decides at the start; tail sampling decides after the outcome is known. See Sampling.
A read-only, point-in-time copy of a dashboard at a public URL — the one door for sharing outside your organization. See Dashboards.
One operation within a trace, with a duration, a status and attributes. Nesting is causality. See Traces.
A check run against your endpoints from outside — HTTP, SSL, DNS, TCP or multi-step. The only signal that exists when your code does not run. See Synthetic monitoring.
One request’s journey across services, made of spans linked by a shared trace_id propagated on every hop.
The W3C header carrying trace context between services. Its absence is why a trace has a hole in it.

​
Next

Signals and the data model

How these fit together.

FAQ

Short answers to common questions.

Platform architecture

The machinery behind the words.