AI Observability Page & Improved RUM Session Search

AI Observability | Monitor LLM Calls and Agent Workflows

AI Observability now has its own page in groundcover. Monitor LLM calls and agent workflows where you already monitor everything else - no new tool, no data leaving your cloud.

groundcover's sensor auto-detects AI API calls via eBPF, capturing model, tokens, cost, and latency for OpenAI, Anthropic, and Bedrock without SDKs or code changes. For deeper visibility into reasoning chains, tool calls, and orchestration flows, instrument with OpenTelemetry GenAI Semantic Conventions and send via OTLP.

The page gives you full trace visibility with cost and token breakdowns, so you can pinpoint slow or expensive calls, understand how agents interact with your production environment, and set monitors to catch regressions. A demo environment is available so you can explore the experience before connecting your own data.

New Features & Improvements

• AI Observability | Launched the AI Observability page with demo environment, enrichment data in Raw tab, cost/token tooltips, and polished drawer UI
• Agent | Redesigned cursor job card, updated input placeholder, added "Add to context" entry points and "Add to Dashboard" from chat, and dismissible suggestions
• RUM | Improved session search with better default search keys including user properties and session metadata
• Monitors | Added label autocomplete for group-by functions and support for multiple connected apps
• Data Explorer | Enabled synced crosshair by default

Bug Fixes

• Traces | Fixed drawer rendering empty when any span had null duration
• Dashboards | Fixed queries dropping arithmetic outside range vector and copying from preview omitting full definition
• Agent | Fixed issues with @mentions, pasted URLs, closing the issue drawer, and input field focus
• Synthetics | Fixed monitor not firing despite identical query
• RUM | Fixed session navigation with arrow keys

‍