Generative AI

as of v1.44.0

Model calls, agents, tools and A2A. Split into its own repository in v1.44.0 and still unreleased.

attributes: 108
metrics: 24
signals: 25
stable: 0
deprecated: 12

Signals

What this domain emits, and the attributes each one carries, most strongly required first. An attribute marked * is required on that signal; * is conditionally required.

Spans 17

a2a.clientclientDevelopmentThis span describes an A2A call from the client side.
a2a.method.name*a2a.message.id*a2a.task.id*a2a.task.state*a2a.tenant*gen_ai.agent.description*gen_ai.agent.name*gen_ai.agent.version*gen_ai.conversation.id*a2a.message.reference_task_idsa2a.protocol.versionerror.type+2 more
a2a.serverserverDevelopmentThis span describes processing of an A2A request by the server.
a2a.method.name*a2a.message.id*a2a.task.id*a2a.task.state*a2a.tenant*gen_ai.agent.description*gen_ai.agent.name*gen_ai.agent.version*gen_ai.conversation.id*a2a.message.reference_task_idsa2a.protocol.versionclient.address+4 more
gen_ai.client.embeddingsclientDevelopmentDescribes a GenAI embeddings operation - a client call to a model or service that returns embeddings for one or more inp
gen_ai.provider.name*error.typegen_ai.embeddings.dimension.countgen_ai.operation.namegen_ai.request.encoding_formatsgen_ai.request.modelgen_ai.response.modelgen_ai.usage.input_tokensserver.addressserver.port
gen_ai.client.inferenceclientDevelopmentThis span represents a logical client call to Generative AI model or service that generates a response or requests a too
gen_ai.provider.name*gen_ai.conversation.id*gen_ai.output.type*gen_ai.prompt.name*gen_ai.prompt.version*gen_ai.request.choice.count*gen_ai.request.seed*gen_ai.request.stream*gen_ai.request.top_k*error.typegen_ai.conversation.compactedgen_ai.operation.name+38 more
gen_ai.create_agent.clientclientDevelopmentDescribes GenAI agent creation and is usually applicable when working with remote agent services.
gen_ai.provider.name*gen_ai.agent.description*gen_ai.agent.id*gen_ai.agent.name*gen_ai.agent.version*error.typegen_ai.operation.namegen_ai.request.modelserver.addressserver.portgen_ai.system_instructions
gen_ai.execute_tool.internalinternalDevelopmentDescribes tool execution span.
gen_ai.operation.name*gen_ai.conversation.id*gen_ai.tool.call.id*error.typegen_ai.agent.namegen_ai.tool.descriptiongen_ai.tool.namegen_ai.tool.typegen_ai.tool.call.argumentsgen_ai.tool.call.result
gen_ai.fetch_response.clientclientDevelopmentDescribes a request to a Generative AI provider that fetches a previously generated model response by its identifier.
gen_ai.operation.name*gen_ai.provider.name*gen_ai.response.id*gen_ai.request.stream_cursor*error.typegen_ai.response.finish_reasonsgen_ai.response.modelgen_ai.response.statusserver.addressserver.portgen_ai.output.messagesgen_ai.system_instructions+1 more
gen_ai.invoke_agent.clientclientDevelopmentDescribes GenAI agent invocation over a remote service.
gen_ai.provider.name*gen_ai.agent.description*gen_ai.agent.id*gen_ai.agent.version*gen_ai.conversation.id*gen_ai.data_source.id*gen_ai.output.type*gen_ai.request.choice.count*gen_ai.request.seed*error.typegen_ai.agent.namegen_ai.operation.name+32 more
gen_ai.invoke_agent.internalinternalDevelopmentDescribes GenAI agent invocation within the same process.
gen_ai.agent.description*gen_ai.conversation.id*gen_ai.data_source.id*gen_ai.output.type*gen_ai.request.choice.count*gen_ai.request.seed*error.typegen_ai.agent.namegen_ai.operation.namegen_ai.request.frequency_penaltygen_ai.request.max_tokensgen_ai.request.model+15 more
gen_ai.invoke_workflow.internalinternalDevelopmentRepresents an operation that executes a coordinated process composed of multiple agents or other operations involving ge
gen_ai.operation.name*gen_ai.conversation.id*gen_ai.workflow.name*error.typegen_ai.input.messagesgen_ai.output.messages
gen_ai.memory.clientclientDevelopmentDescribes a memory operation, including memory store lifecycle operations and memory record operations.
gen_ai.operation.name*gen_ai.memory.record.id*gen_ai.memory.store.id*gen_ai.provider.name*error.typegen_ai.memory.record.countserver.addressserver.portgen_ai.memory.query.textgen_ai.memory.records
gen_ai.plan.internalinternalDevelopmentRepresents an agent planning or task decomposition phase.
gen_ai.operation.name*gen_ai.agent.name*error.type
gen_ai.retrieval.clientclientDevelopmentDescribes GenAI retrieval span - a request to a Generative AI service or framework that retrieves relevant information o
gen_ai.operation.name*gen_ai.data_source.id*gen_ai.provider.name*gen_ai.request.model*error.typegen_ai.retrieval.top_kserver.addressserver.portgen_ai.retrieval.documentsgen_ai.retrieval.query.text
anthropic.inference.clientclientDeprecatedSemantic Conventions for Anthropic client spans extend and override the semantic conventions for Gen AI Spans.
error.type*server.port*gen_ai.conversation.idgen_ai.input.messagesgen_ai.operation.namegen_ai.output.messagesgen_ai.output.typegen_ai.request.choice.countgen_ai.request.frequency_penaltygen_ai.request.max_tokensgen_ai.request.modelgen_ai.request.presence_penalty+17 more
gen_ai.embeddings.clientclientDeprecatedDescribes GenAI embeddings span - a request to a Generative AI model or service that generates an embeddings based on th
error.type*gen_ai.embeddings.dimension.countgen_ai.operation.namegen_ai.provider.namegen_ai.request.encoding_formatsgen_ai.request.modelgen_ai.response.modelgen_ai.usage.input_tokensserver.addressserver.port
gen_ai.inference.clientclientDeprecatedThis span represents a client call to Generative AI model or service that generates a response or requests a tool call b
error.type*gen_ai.conversation.idgen_ai.input.messagesgen_ai.operation.namegen_ai.output.messagesgen_ai.output.typegen_ai.provider.namegen_ai.request.choice.countgen_ai.request.frequency_penaltygen_ai.request.max_tokensgen_ai.request.modelgen_ai.request.presence_penalty+19 more
openai.inference.clientclientDeprecatedSemantic Conventions for OpenAI client spans extend and override the semantic conventions for Gen AI Spans.
error.type*gen_ai.conversation.idgen_ai.input.messagesgen_ai.operation.namegen_ai.output.messagesgen_ai.output.typegen_ai.request.choice.countgen_ai.request.frequency_penaltygen_ai.request.max_tokensgen_ai.request.modelgen_ai.request.presence_penaltygen_ai.request.seed+21 more

Events 8

gen_ai.assistant.messageDeprecatedThis event describes the assistant message passed to GenAI system.
gen_ai.system
gen_ai.choiceDeprecatedThis event describes the Gen AI response message.
gen_ai.system
gen_ai.client.inference.operation.detailsDevelopmentDescribes the details of a GenAI completion request including chat history and parameters.
gen_ai.client.operation.exceptionDevelopmentThis event represents an exception that occurred during a Generative AI client operation, such as API errors, rate limit
exception.message*exception.type*exception.stacktrace
gen_ai.evaluation.resultDevelopmentThis event captures the result of evaluating GenAI output for quality, accuracy, or other characteristics. This event SH
gen_ai.evaluation.name*error.type*gen_ai.evaluation.score.label*gen_ai.evaluation.score.value*gen_ai.evaluation.explanationgen_ai.response.id
gen_ai.system.messageDeprecatedThis event describes the system instructions passed to the GenAI model.
gen_ai.system
gen_ai.tool.messageDeprecatedThis event describes the response from a tool or function call passed to the GenAI model.
gen_ai.system
gen_ai.user.messageDeprecatedThis event describes the user message passed to the GenAI model.
gen_ai.system

Metrics 24

gen_ai.client.embeddings.duration
Developmenthistogram · sGenAI client embeddings operation duration.
gen_ai.client.embeddings.operation.input_tokens
Developmenthistogram · {token}The number of input tokens used per embeddings operation.
gen_ai.client.inference.duration
Developmenthistogram · sGenAI client inference operation duration.
gen_ai.client.inference.operation.input_tokens
Developmenthistogram · {token}The number of input (prompt) tokens used per inference operation.
gen_ai.client.inference.operation.output_tokens
Developmenthistogram · {token}The number of output (completion) tokens used per inference operation.
gen_ai.client.inference.time_per_output_chunk
Developmenthistogram · sTime per output chunk, recorded for each chunk received after the first one, measured as the time elapsed from the end of the previous chunk
gen_ai.client.inference.time_to_first_chunk
Developmenthistogram · sTime from the start of a logical inference operation to the first chunk in its streaming response.
gen_ai.client.inference.usage.cache_read.input_tokens
Developmentcounter · {token}The number of input tokens served from a provider-managed cache.
gen_ai.client.inference.usage.cache_write.input_tokens
Developmentcounter · {token}The number of input tokens written to a provider-managed cache.
gen_ai.client.inference.usage.input_tokens
Developmentcounter · {token}The number of input (prompt) tokens used, including cached tokens.
gen_ai.client.inference.usage.output_tokens
Developmentcounter · {token}The number of output (completion) tokens used, including reasoning tokens.
gen_ai.client.inference.usage.reasoning.output_tokens
Developmentcounter · {token}The number of output tokens used for reasoning (e.g. chain-of-thought, extended thinking).
gen_ai.client.operation.duration
Developmenthistogram · sGenAI operation duration.
gen_ai.client.operation.time_per_output_chunk
Deprecatedhistogram · sTime per output chunk, recorded for each chunk received after the first one, measured as the time elapsed from the end of the previous chunk
gen_ai.client.operation.time_to_first_chunk
Deprecatedhistogram · sTime to receive the first chunk, measured from when the client issues the generation request to when the first chunk is received in the resp
gen_ai.client.token.usage
Deprecatedhistogram · {token}Number of input and output tokens used.
gen_ai.execute_tool.duration
Developmenthistogram · sThe duration of a single tool execution.
gen_ai.invoke_agent.duration
Developmenthistogram · sThe end-to-end duration of a single in-process agent invocation, from the moment the invocation starts until the agent emits the last chunk
gen_ai.invoke_agent.inference_calls
Developmenthistogram · {inference_call}The number of inference (model) calls a GenAI agent makes during a single invocation.
gen_ai.invoke_agent.tool_calls
Developmenthistogram · {tool_call}The number of tool calls a GenAI agent makes during a single invocation.
gen_ai.invoke_workflow.duration
Developmenthistogram · sRecords duration of GenAI workflow.
gen_ai.server.request.duration
Developmenthistogram · sGenerative AI server request duration such as time-to-last byte or last output token.
gen_ai.server.time_per_output_token
Developmenthistogram · sTime per output token generated after the first token for successful responses.
gen_ai.server.time_to_first_token
Developmenthistogram · sTime to generate first token for successful responses.

Specification

Spansare bound by the trace specification — 302 requirements across 7 documents, 168 of them MUST or SHALL.
Metricsare bound by the metrics specification — 561 requirements across 8 documents, 287 of them MUST or SHALL.
Eventsare bound by the logs specification — 162 requirements across 5 documents, 83 of them MUST or SHALL.

The specification prose never names this domain. That is not an omission: the specification defines signals, not subject matter — what binds gen_ai telemetry is the general rules above, and the conventions themselves carry the rest.

On the wire, this lands in trace/v1/trace, metrics/v1/metrics, logs/v1/logs. Field-by-field definitions are under OTLP.

Attributes 108

gen_ai97 attributes

gen_ai.agent.description
DevelopmentGenAIFree-form description of the GenAI agent provided by the application.
gen_ai.agent.id
DevelopmentGenAIThe unique and stable identifier of the GenAI hosted agent resource.
gen_ai.agent.name
DevelopmentGenAIHuman-readable name of the GenAI agent provided by the application.
gen_ai.agent.version
DevelopmentGenAIThe version of the GenAI agent.
gen_ai.completion
DeprecatedDeprecated, use Event API to report completions contents.
gen_ai.conversation.compacted
DevelopmentGenAIIndicates whether the effective conversation context used for this operation is a compacted view of a prior conversation.
gen_ai.conversation.id
DevelopmentGenAIThe unique identifier for a conversation (session, thread), used to store and correlate messages within this conversation.
gen_ai.data_source.id
DevelopmentGenAIThe data source identifier.
gen_ai.embeddings.dimension.count
DevelopmentGenAIThe number of dimensions the resulting output embeddings should have.
gen_ai.evaluation.explanation
DevelopmentGenAIA free-form explanation for the assigned score provided by the evaluator.
gen_ai.evaluation.name
DevelopmentGenAIThe name of the evaluation metric used for the GenAI response.
gen_ai.evaluation.score.label
DevelopmentGenAIHuman readable label for evaluation.
gen_ai.evaluation.score.value
DevelopmentGenAIThe evaluation score returned by the evaluator.
gen_ai.input.messages
DevelopmentGenAIThe chat history provided to the model as an input.
gen_ai.input.messages_ref
DevelopmentGenAIAn absolute URI identifying the input messages in external storage.
gen_ai.main_agent.description
DevelopmentGenAIThe free-form description of the top-level Generative AI agent in the process.
gen_ai.main_agent.id
DevelopmentGenAIThe unique and stable identifier of the top-level Generative AI agent in the process
gen_ai.main_agent.name
DevelopmentGenAIThe human-readable name of the top-level Generative AI agent in the process.
gen_ai.memory.query.text
DevelopmentGenAIThe search query used to retrieve memories.
gen_ai.memory.record.count
DevelopmentGenAIThe number of memory records relevant to the operation.
gen_ai.memory.record.id
DevelopmentGenAIThe unique identifier of the memory record.
gen_ai.memory.records
DevelopmentGenAIThe memory records stored or retrieved in a memory operation.
gen_ai.memory.store.id
DevelopmentGenAIThe unique identifier of the memory store.
gen_ai.openai.request.response_format
DeprecatedDeprecated, use gen_ai.output.type.
gen_ai.openai.request.seed
DeprecatedDeprecated, use gen_ai.request.seed.
gen_ai.openai.request.service_tier
DeprecatedDeprecated, use openai.request.service_tier.
gen_ai.openai.response.service_tier
DeprecatedDeprecated, use openai.response.service_tier.
gen_ai.openai.response.system_fingerprint
DeprecatedDeprecated, use openai.response.system_fingerprint.
gen_ai.operation.name
DevelopmentGenAIThe name of the operation being performed.
gen_ai.output.messages
DevelopmentGenAIMessages returned by the model where each message represents a specific model response (choice, candidate).
gen_ai.output.messages_ref
DevelopmentGenAIAn absolute URI identifying the output messages in external storage.
gen_ai.output.type
DevelopmentGenAIRepresents the content type requested by the client.
gen_ai.prompt
DeprecatedDeprecated, use Event API to report prompt contents.
gen_ai.prompt.name
DevelopmentGenAIThe name of the prompt that uniquely identifies it.
gen_ai.prompt.variable
DevelopmentGenAIThe variables supplied to the prompt template, the <key> being the variable name, the value being the variable value.
gen_ai.prompt.version
DevelopmentGenAIThe version of the prompt template used.
gen_ai.provider.name
DevelopmentGenAIThe Generative AI provider as identified by the client or server instrumentation.
gen_ai.request.choice.count
DevelopmentGenAIThe target number of candidate completions to return.
gen_ai.request.encoding_formats
DevelopmentGenAIThe encoding formats requested in an embeddings operation, if specified.
gen_ai.request.frequency_penalty
DevelopmentGenAIThe frequency penalty setting for the GenAI request.
gen_ai.request.max_tokens
DevelopmentGenAIThe maximum number of tokens the model generates for a request.
gen_ai.request.model
DevelopmentGenAIThe name of the GenAI model a request is being made to.
gen_ai.request.presence_penalty
DevelopmentGenAIThe presence penalty setting for the GenAI request.
gen_ai.request.previous_response.id
DevelopmentGenAIThe unique identifier of a previous response or interaction used to provide context for the current operation.
gen_ai.request.reasoning.level
DevelopmentGenAIThe reasoning or thinking effort level requested for a GenAI model.
gen_ai.request.seed
DevelopmentGenAIRequests with same seed value more likely to return same result.
gen_ai.request.stop_sequences
DevelopmentGenAIList of sequences that the model will use to stop generating further tokens.
gen_ai.request.stream
DevelopmentGenAIIndicates whether the GenAI request was made in streaming mode.
gen_ai.request.stream_cursor
DevelopmentGenAIThe cursor identifying the last streamed event already received, used to resume a streamed response from that position.
gen_ai.request.temperature
DevelopmentGenAIThe temperature setting for the GenAI request.
gen_ai.request.top_k
DevelopmentGenAIThe top-K sampling setting for the GenAI request: restricts token generation at each step to the K most likely next tokens.
gen_ai.request.top_p
DevelopmentGenAIThe top_p sampling setting for the GenAI request.
gen_ai.response.finish_reasons
DevelopmentGenAIArray of reasons the model stopped generating tokens, corresponding to each generation received.
gen_ai.response.id
DevelopmentGenAIThe unique identifier for the completion.
gen_ai.response.model
DevelopmentGenAIThe name of the model that generated the response.
gen_ai.response.status
DevelopmentGenAIThe lifecycle status of a generated response, as reported by the provider when the response is fetched or polled.
gen_ai.response.time_to_first_chunk
DevelopmentGenAITime from the start of a logical inference operation to the first chunk in its streaming response, in seconds.
gen_ai.retrieval.documents
DevelopmentGenAIThe documents retrieved.
gen_ai.retrieval.query.text
DevelopmentGenAIThe query text used for retrieval.
gen_ai.retrieval.top_k
DevelopmentGenAIThe maximum number of documents the retriever was asked to return for the query (also known as k, limit, or max_num_results).
gen_ai.skill.description
DevelopmentGenAIThe description of the Agent Skill.
gen_ai.skill.name
DevelopmentGenAIThe name of the Agent Skill.
gen_ai.skill.resource.name
DevelopmentGenAIThe skill-relative name of the resource (script, reference, or asset) being accessed or executed.
gen_ai.skill.source.uri
DevelopmentGenAIThe source URI for loading the skill.
gen_ai.system
DeprecatedDeprecated, use gen_ai.provider.name instead.
gen_ai.system_instructions
DevelopmentGenAIThe system message or instructions provided to the GenAI model separately from the chat history.
gen_ai.system_instructions_ref
DevelopmentGenAIAn absolute URI identifying the system instructions in external storage.
gen_ai.token.modality
DevelopmentGenAIThe modality of the tokens being counted.
gen_ai.token.type
DeprecatedThe type of token being counted.
gen_ai.tool.call.arguments
DevelopmentGenAIParameters passed to the tool call.
gen_ai.tool.call.id
DevelopmentGenAIThe tool call identifier supplied by the model or provider.
gen_ai.tool.call.result
DevelopmentGenAIThe result returned by the tool call (if any and if execution was successful).
gen_ai.tool.definitions
DevelopmentGenAIThe list of tool definitions available to the GenAI agent or model.
gen_ai.tool.definitions_ref
DevelopmentGenAIAn absolute URI identifying the tool definitions in external storage.
gen_ai.tool.description
DevelopmentGenAIThe tool description.
gen_ai.tool.name
DevelopmentGenAIName of the tool utilized by the agent.
gen_ai.tool.type
DevelopmentGenAIType of the tool utilized by the agent
gen_ai.transfer.mode
DevelopmentGenAIDescribes how control passes to the transfer target.
gen_ai.transfer.target.name
DevelopmentGenAIThe human-readable name of the agent targeted by the transfer.
gen_ai.usage.audio.cache_read.input_tokens
DevelopmentGenAIThe number of audio input tokens served from a provider-managed cache.
gen_ai.usage.audio.input_tokens
DevelopmentGenAIThe number of audio input tokens.
gen_ai.usage.audio.output_tokens
DevelopmentGenAIThe number of audio output tokens.
gen_ai.usage.cache_creation.input_tokens
DeprecatedThe number of input tokens written to a provider-managed cache.
gen_ai.usage.cache_read.input_tokens
DevelopmentGenAIThe number of input tokens served from a provider-managed cache.
gen_ai.usage.cache_write.input_tokens
DevelopmentGenAIThe number of input tokens written to a provider-managed cache.
gen_ai.usage.completion_tokens
DeprecatedDeprecated, use gen_ai.usage.output_tokens instead.
gen_ai.usage.image.cache_read.input_tokens
DevelopmentGenAIThe number of image input tokens served from a provider-managed cache.
gen_ai.usage.image.input_tokens
DevelopmentGenAIThe number of image input tokens.
gen_ai.usage.image.output_tokens
DevelopmentGenAIThe number of image output tokens.
gen_ai.usage.input_tokens
DevelopmentGenAIThe number of tokens used in the GenAI input (prompt).
gen_ai.usage.output_tokens
DevelopmentGenAIThe number of tokens used in the GenAI response (completion).
gen_ai.usage.prompt_tokens
DeprecatedDeprecated, use gen_ai.usage.input_tokens instead.
gen_ai.usage.reasoning.output_tokens
DevelopmentGenAIThe number of output tokens used for reasoning (e.g. chain-of-thought, extended thinking).
gen_ai.usage.text.cache_read.input_tokens
DevelopmentGenAIThe number of text input tokens served from a provider-managed cache.
gen_ai.usage.text.input_tokens
DevelopmentGenAIThe number of text input tokens.
gen_ai.usage.text.output_tokens
DevelopmentGenAIThe number of text output tokens.
gen_ai.workflow.name
DevelopmentGenAIHuman-readable name of the GenAI workflow provided by the application.

a2a7 attributes

a2a.message.id
DevelopmentGenAIThe unique identifier of the A2A message.
a2a.message.reference_task_ids
DevelopmentGenAITask IDs referenced by the A2A request message.
a2a.method.name
DevelopmentGenAIThe name of the A2A request method.
a2a.protocol.version
DevelopmentGenAIThe version of the A2A protocol used.
a2a.task.id
DevelopmentGenAIThe unique identifier of an A2A task.
a2a.task.state
DevelopmentGenAIThe state of an A2A task.
a2a.tenant
DevelopmentGenAIThe opaque routing identifier of the target A2A tenant.

openai4 attributes

openai.api.type
DevelopmentGenAIThe type of OpenAI API being used.
openai.request.service_tier
DevelopmentGenAIThe service tier requested. May be a specific tier, default, or auto.
openai.response.service_tier
DevelopmentGenAIThe service tier used for the response.
openai.response.system_fingerprint
DevelopmentGenAIA fingerprint to track any eventual change in the Generative AI environment.

Recent changes

Breaking and notable changes to this domain across the last 6 semantic-conventions releases.

deprecatedattributegen_ai.agent.descriptionMoved to the OpenTelemetry GenAI semantic conventions repository.
deprecatedattributegen_ai.agent.idMoved to the OpenTelemetry GenAI semantic conventions repository.
deprecatedattributegen_ai.agent.nameMoved to the OpenTelemetry GenAI semantic conventions repository.
deprecatedattributegen_ai.agent.versionMoved to the OpenTelemetry GenAI semantic conventions repository.
deprecatedmetricgen_ai.client.operation.durationMoved to the OpenTelemetry GenAI semantic conventions repository.
deprecatedmetricgen_ai.client.operation.time_per_output_chunkMoved to the OpenTelemetry GenAI semantic conventions repository.
deprecatedmetricgen_ai.client.operation.time_to_first_chunkMoved to the OpenTelemetry GenAI semantic conventions repository.
deprecatedmetricgen_ai.client.token.usageMoved to the OpenTelemetry GenAI semantic conventions repository.
deprecatedattributegen_ai.conversation.idMoved to the OpenTelemetry GenAI semantic conventions repository.
deprecatedattributegen_ai.data_source.idMoved to the OpenTelemetry GenAI semantic conventions repository.

59 more in the full release history

Topics are this site's grouping of upstream namespaces, not a structure the OpenTelemetry project publishes. JSON · All domains