What problem does it solve? Teams running the Bubbles agent orchestration framework need a uniform, swappable way to fetch live production telemetry (alerts, SLO burn, error rate, deploy impact) from their own monitoring stack, and a method to mine validate-plane traces for hidden defects during live-category tests. ## Core Features & Use Cases - Uniform 4-verb adapter contract: Every adapter under bubbles/adapters/observability/ implements fetch-alerts, fetch-slo-burn, fetch-error-rate, and fetch-deploy-impact with defined JSON output shapes and exit-code semantics, verified by observability-adapter-lint.sh. - Two-plane wiring: Declarative (plane, signal) → {adapter, profile} resolution in traceContracts.observability.endpoints splits the ephemeral validate test stack from read-only prod operate-plane consumers (bubbles.stabilize, bubbles.upkeep, bubbles.train). - Trace-driven defect discovery: During integration/e2e/stress/load tests against a wired stack, mine telemetry for error spans, latency outliers, fan-out/N+1 patterns, and missing spans, then file findings through bubbles.bug. - Use Case: You run Grafana Cloud Prometheus and want Bubbles ops agents to read live SLO burn and alerts; author a prometheus.sh-style adapter, wire it in .github/bubbles-project.yaml, and incident diagnosis plus promote/rollback gating light up immediately. ## Quick Start Ask the agent to author a new observability adapter for your telemetry stack following the four-verb contract and wire it into the validate and operate planes of your project's traceContracts configuration.
