13 articles tagged
Always-on tracing went to beta on 21 August 2026 at $0.50 per 1M span units, a price only the changelog states. Trace retention stops at 3 days even where Observability Plus keeps 30 days of logs.
Version 1.0.0 of the anthropic Python package landed on 20 August 2026 and moved the HTTP layer to httpx2. The upgrade installs, the type checks pass, and the mocks quietly stop matching.
Cloudflare restored WebSocket data transfer accounting on 14 August 2026 but never dated the affected period. EdgeResponseBytes counts one direction; the websocket_analytics Logpush dataset counts four.
AWS gave Aurora DSQL a CloudWatch Database Insights metric on 20 August 2026. The mode model does not list DSQL and the AWS Price List API carries no Aurora DSQL Database Insights SKU.
AWS shipped log group tag propagation for CloudWatch Centralization on 19 August 2026. The opt-in carries a rename requirement, a non-adjustable 5 TPS tagging ceiling, and a health field that stays green.
Vercel's new metric() function emits custom metrics from Vercel Functions. Every data point counts as one Observability event at $1.20 per million, with hard caps of 100 metrics per invocation and 50 attributes each.
Gemini Enterprise now traces an agent call end to end, from the assistant prompt through the orchestration layer to the third-party API. Two new spans, a 30-day retention window and a $0.20 per million span meter decide
Grafana Cloud added Agent Observability for LLM agents, built on OpenTelemetry. This guide covers instrumentation with OpenLIT, what the traces show, and the trace-volume cost math versus Datadog, Langfuse and Arize
Amazon Bedrock AgentCore moved every agent trace, prompt and log into a single per-agent CloudWatch log group and lifted default limits to 5,000 concurrent sessions. A config-level guide to the July 2026 observability
CloudWatch Coding Agent Insights, launched July 20, 2026, gives engineering leaders OpenTelemetry dashboards for AI coding tools: how to measure token spend, per-team cost and developer productivity across Claude Code
Vendor posts say OpenTelemetry's GenAI client spans went stable in early 2026. The specification says Development on every page, which sits below Alpha. Here is what that means for your instrumentation.
Agents do not crash. They return a confident, plausible, wrong answer, and nothing goes red. Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027. Here is what to instrument, what to score
Most enterprise AI agents never reach production, and quality kills a third that do. Five hard-won lessons from shipping production AI agents: scope, harness, observability, guardrails and human oversight.
