observability-llm-obs

You can monitor LLM token usage and cost by analyzing ingested trace and metrics data streams to track token utilization and estimate LLM call expenses. This Skill leverages ES|QL and OpenTelemetry GenAI semantic conventions to query and calculate costs.

Track LLM performance and latency by analyzing data ingested into Elastic to measure throughput, error rates, and response delays for LLM operations. It uses Elasticsearch APIs and Kibana APIs to query trace data streams for detailed performance insights.

Yes, you can analyze agentic workflow orchestration and call chaining with Elastic to understand execution flow and identify bottlenecks. This Skill queries ingested trace data to map out complex LLM operations and workflow sequences.

The best way to identify response quality issues in LLM operations is to analyze ingested metrics data for content filtering events and errors. This Skill uses ES|QL queries to pinpoint specific LLM calls experiencing degraded response quality.

Yes, this LLM observability approach relies on OpenTelemetry GenAI semantic conventions and integration-specific fields to query trace and metrics data streams. This standardization allows accurate extraction of performance and cost data.