genai:semantic-conventions

You standardize LLM observability by applying OpenTelemetry semantic conventions to define `gen_ai.*` attributes for requests, responses, and conversation context. This ensures consistent logging of AI interactions across different frameworks.

Yes, you can log OpenAI token usage without custom mapping by using the `opentelemetry-instrumentation-openai` dependency. It automatically instruments OpenAI SDK calls and logs model parameters to your observability backend.

Yes, OpenTelemetry instrumentation works with Phoenix and MLflow by transforming `gen_ai.*` attributes into OpenInference and MLflow formats. This cross-platform compatibility eliminates the need for custom mapping logic.

The best way to instrument AI agents for consistent logging is using standardized OpenTelemetry semantic conventions. This facilitates tracking model calls, token usage, and conversation IDs uniformly across various AI frameworks and backends.

Yes, you need Python to set up OpenTelemetry semantic conventions for GenAI. The auto-instrumentation and manual span attribute setting require Python along with the `opentelemetry-sdk` and `opentelemetry-exporter-otlp` packages.

OpenTelemetry exports miss GenAI conversation context if you do not manually set span attributes. You must explicitly define `gen_ai.*` attributes for requests and responses to ensure meaningful observability data for LLM operations.