So far we put all the messages including the system prompt on gen_ai.request.messages. Instead, the system prompt should be pulled out and set as a separate attribute called gen_ai.system_instructions. For now we will search for the first system message and use that.

The attribute should adhere to the structure given in the otel spec: https://opentelemetry.io/docs/specs/semconv/registry/attributes/gen-ai/#gen-ai-system-instructions