GenAI conventions2026-09-22

22 Sept 2026since 2026-09-16

0 breaking2 notable20 editorial

From the upstream commit

Fix usage metrics to provide meaningful aggregation and break down by modality, reasoning, cache usage

Read the commit

Notable2

removedmetricgen_ai.client.token.usageRemoved from the registry (was development).
removedattributegen_ai.token.typeRemoved from the registry (was development).

Editorial20

usageattributegen_ai.operation.nameNow referenced by metric_attributes.gen_ai.inference_usage, gen_ai.client.inference.operation.input_tokens and 6 others (required).
usageattributegen_ai.operation.nameNo longer referenced by gen_ai.client.token.usage.
usageattributegen_ai.provider.nameNow referenced by metric_attributes.gen_ai.inference_usage, gen_ai.client.inference.operation.input_tokens and 6 others (required).
usageattributegen_ai.provider.nameNo longer referenced by gen_ai.client.token.usage.
usageattributegen_ai.request.modelNow referenced by metric_attributes.gen_ai.inference_usage, gen_ai.client.inference.operation.input_tokens and 6 others (conditionally required).
usageattributegen_ai.request.modelNo longer referenced by gen_ai.client.token.usage.
usageattributegen_ai.response.modelNow referenced by metric_attributes.gen_ai.inference_usage, gen_ai.client.inference.operation.input_tokens and 6 others (recommended).
usageattributegen_ai.response.modelNo longer referenced by gen_ai.client.token.usage.
usageattributeopenai.response.service_tierNow referenced by gen_ai.client.inference.operation.input_tokens, gen_ai.client.inference.operation.output_tokens and 5 others (recommended).
usageattributeopenai.response.service_tierNo longer referenced by gen_ai.client.token.usage.
usageattributeopenai.response.system_fingerprintNow referenced by gen_ai.client.inference.operation.input_tokens, gen_ai.client.inference.operation.output_tokens and 5 others (recommended).
usageattributeopenai.response.system_fingerprintNo longer referenced by gen_ai.client.token.usage.
addedmetricgen_ai.client.inference.operation.input_tokensThe number of input (prompt) tokens used per inference operation.
addedmetricgen_ai.client.inference.operation.output_tokensThe number of output (completion) tokens used per inference operation.
addedmetricgen_ai.client.inference.usage.cache_read.input_tokensThe number of input tokens served from a provider-managed cache.
addedmetricgen_ai.client.inference.usage.cache_write.input_tokensThe number of input tokens written to a provider-managed cache.
addedmetricgen_ai.client.inference.usage.input_tokensThe number of input (prompt) tokens used, including cached tokens.
addedmetricgen_ai.client.inference.usage.output_tokensThe number of output (completion) tokens used, including reasoning tokens.
addedmetricgen_ai.client.inference.usage.reasoning.output_tokensThe number of output tokens used for reasoning (e.g. chain-of-thought, extended thinking).
addedattributegen_ai.token.modalityThe modality of the tokens being counted.