GenAI conventions2026-10-06

6 Oct 2026since 2026-10-05.2

0 breaking3 notable103 editorial

From the upstream commit

Use gen_ai.client.inference.duration metric instead of gen_ai.client.operation.duration and move inference signals to one file

Read the commit

Notable3

removedmetricgen_ai.client.operation.time_per_output_chunkRemoved from the registry (was development).
removedmetricgen_ai.client.operation.time_to_first_chunkRemoved from the registry (was development).
removedsignalgen_ai.inference.clientRemoved from the registry (was development).

Editorial103

usageattributeaws.bedrock.guardrail.idNow referenced by gen_ai.client.inference (required).
usageattributeaws.bedrock.guardrail.idNo longer referenced by gen_ai.inference.client.
usageattributeaws.bedrock.knowledge_base.idNow referenced by gen_ai.client.inference (recommended).
usageattributeaws.bedrock.knowledge_base.idNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.conversation.compactedNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.conversation.compactedNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.conversation.idNow referenced by gen_ai.client.inference (conditionally required).
usageattributegen_ai.conversation.idNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.input.messagesNow referenced by gen_ai.client.inference (opt in).
usageattributegen_ai.input.messagesNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.operation.nameNow referenced by gen_ai.client.inference, gen_ai.client.inference.duration and 2 others (recommended).
usageattributegen_ai.operation.nameNo longer referenced by gen_ai.inference.client, gen_ai.client.operation.time_per_output_chunk and 1 other.
usageattributegen_ai.output.messagesNow referenced by gen_ai.client.inference (opt in).
usageattributegen_ai.output.messagesNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.output.typeNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.output.typeNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.prompt.nameNow referenced by gen_ai.client.inference (conditionally required).
usageattributegen_ai.prompt.nameNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.prompt.variableNow referenced by gen_ai.client.inference (opt in).
usageattributegen_ai.prompt.variableNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.prompt.versionNow referenced by gen_ai.client.inference (conditionally required).
usageattributegen_ai.prompt.versionNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.provider.nameNow referenced by gen_ai.client.inference, gen_ai.client.inference.duration and 2 others (required).
usageattributegen_ai.provider.nameNo longer referenced by gen_ai.inference.client, gen_ai.client.operation.time_per_output_chunk and 1 other.
usageattributegen_ai.request.choice.countNow referenced by gen_ai.client.inference (conditionally required).
usageattributegen_ai.request.choice.countNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.frequency_penaltyNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.request.frequency_penaltyNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.max_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.request.max_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.modelNow referenced by gen_ai.client.inference, gen_ai.client.inference.duration and 2 others (required).
usageattributegen_ai.request.modelNo longer referenced by gen_ai.inference.client, gen_ai.client.operation.time_per_output_chunk and 1 other.
usageattributegen_ai.request.presence_penaltyNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.request.presence_penaltyNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.previous_response.idNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.request.previous_response.idNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.reasoning.levelNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.request.reasoning.levelNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.seedNow referenced by gen_ai.client.inference (conditionally required).
usageattributegen_ai.request.seedNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.stop_sequencesNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.request.stop_sequencesNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.streamNow referenced by gen_ai.client.inference (conditionally required).
usageattributegen_ai.request.streamNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.temperatureNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.request.temperatureNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.top_kNow referenced by gen_ai.client.inference (conditionally required).
usageattributegen_ai.request.top_kNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.request.top_pNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.request.top_pNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.response.finish_reasonsNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.response.finish_reasonsNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.response.idNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.response.idNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.response.modelNow referenced by gen_ai.client.inference, gen_ai.client.inference.duration and 2 others (recommended).
usageattributegen_ai.response.modelNo longer referenced by gen_ai.inference.client, gen_ai.client.operation.time_per_output_chunk and 1 other.
usageattributegen_ai.response.time_to_first_chunkNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.response.time_to_first_chunkNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.system_instructionsNow referenced by gen_ai.client.inference (opt in).
usageattributegen_ai.system_instructionsNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.tool.definitionsNow referenced by gen_ai.client.inference (opt in).
usageattributegen_ai.tool.definitionsNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.audio.cache_read.input_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.audio.cache_read.input_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.audio.input_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.audio.input_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.audio.output_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.audio.output_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.cache_read.input_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.cache_read.input_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.cache_write.input_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.cache_write.input_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.image.cache_read.input_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.image.cache_read.input_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.image.input_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.image.input_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.image.output_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.image.output_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.input_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.input_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.output_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.output_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.reasoning.output_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.reasoning.output_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.text.cache_read.input_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.text.cache_read.input_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.text.input_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.text.input_tokensNo longer referenced by gen_ai.inference.client.
usageattributegen_ai.usage.text.output_tokensNow referenced by gen_ai.client.inference (recommended).
usageattributegen_ai.usage.text.output_tokensNo longer referenced by gen_ai.inference.client.
usageattributeopenai.api.typeNow referenced by gen_ai.client.inference (recommended).
usageattributeopenai.api.typeNo longer referenced by gen_ai.inference.client.
usageattributeopenai.request.service_tierNow referenced by gen_ai.client.inference (conditionally required).
usageattributeopenai.request.service_tierNo longer referenced by gen_ai.inference.client.
usageattributeopenai.response.service_tierNow referenced by gen_ai.client.inference and gen_ai.client.inference.duration (conditionally required).
usageattributeopenai.response.service_tierNo longer referenced by gen_ai.inference.client.
usageattributeopenai.response.system_fingerprintNow referenced by gen_ai.client.inference and gen_ai.client.inference.duration (recommended).
usageattributeopenai.response.system_fingerprintNo longer referenced by gen_ai.inference.client.
addedsignalgen_ai.client.inferenceThis span represents a client call to Generative AI model or service that generates a response or requests a tool call based on the input prompt.
addedmetricgen_ai.client.inference.durationGenAI client inference operation duration.
addedmetricgen_ai.client.inference.time_per_output_chunkTime per output chunk, recorded for each chunk received after the first one, measured as the time elapsed from the end of the previous chunk to the end of the ...
addedmetricgen_ai.client.inference.time_to_first_chunkTime to receive the first chunk, measured from when the client issues the generation request to when the first chunk is received in the response stream.
guidancemetricgen_ai.client.operation.durationGuidance note changed.