Prompt
OpenTelemetry LLM stack
Latest observation
Oct 9, 2026
OpenAI APIWeb search: off
Could you clarify what you mean by “OpenTelemetry LLM stack”?
A few possible interpretations:
-
Observability stack for LLM apps using OpenTelemetry
- tracing prompts, completions, tool calls, retrieval, latency, cost, tokens, errors
-
A recommended architecture
- OpenTelemetry SDK + collector + backend for logs/metrics/traces for an LLM app
-
Instrumenting a specific LLM framework
- e.g. LangChain, LlamaIndex, OpenAI SDK, Anthropic, vLLM, etc.
-
A reference implementation
- code example showing how to emit spans and metrics from an LLM pipeline
If you want, I can give you any of these:
- high-level architecture
- best practices
- sample code
- tooling comparison
- production deployment setup
If helpful, I can start with a concise reference architecture for an OpenTelemetry-based LLM application stack.