Lessons from Building an OTel Normalizer for GenAI (Part 1)

In the blog post, Anais Dotis from groundcover discusses the challenges and insights gained from developing a GenAI SDK, framework, and provider-agnostic AI observability solution using OpenTelemetry (OTel). The goal was to simplify AI observability akin to existing methods for infrastructure, APM, and RUM data collection through their eBPF sensor. However, the team encountered significant complexities due to inconsistent telemetry attributes across various SDKs, frameworks, and LLM providers, highlighting the disparity between the idealized narrative of a unified OTel standard and the real-world variations and quirks. Groundcover’s solution involves normalizing GenAI spans from multiple sources into a single canonical view to manage differences in model naming, token semantics, cost calculations, and provider names. The post emphasizes the importance of bridging these gaps to facilitate reliable AI observability and promises to delve deeper into specific technical aspects in the subsequent part of the series.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.