waaseyaa/ai-observability

Trace recording, cost tracking, and anomaly detection for Waaseyaa agentic framework

Maintainers

Package info

github.com/waaseyaa/ai-observability

pkg:composer/waaseyaa/ai-observability

Statistics

Installs: 1 219

Dependents: 2

Suggesters: 0

Stars: 0

Open Issues: 0

Security

Advisories: 0

Aikido package health analysis

v0.1.0-alpha.249 2026-06-25 00:50 UTC

Requires

Requires (Dev)

GPL-2.0-or-later edcfb966c07b3ad7207e78caff1b52a8c445ad31

This package is auto-updated.

Last update: 2026-06-25 00:57:02 UTC

README

Cost tracking, budget enforcement, outcome logging, and anomaly detection for the Waaseyaa agentic framework, plus a (currently dormant) trace recorder.

Layer 5 (AI). Zero dependencies on other AI packages.

The cost / outcome / metrics path is wired and live: AgentRunTelemetryListener (subscribed in AgentTelemetryServiceProvider::boot()) consumes the AgentRun* lifecycle events that ai-agent fires (AgentExecutor, RunAgentHandler) and feeds metrics + Telescope; LlmCallListener feeds the cost accountant.

Trace recording is not active yet. TraceRecorderInterface is bound to TraceRecorder (and NullTraceRecorder when disabled), but no production code calls TraceRecorder::startTrace() — nothing opens a trace, so the trace engine is dormant. This is a known functional gap tracked in #1743 for a post-beta pass; it does not affect the cost/outcome telemetry above.

See the design at docs/history/superpowers/specs/2026-04-14-ai-observability-design.md.