opentelemetry-go/semconv/v1.41.0/genaiconv/metric.go

// Code generated from semantic convention specification. DO NOT EDIT.

// Copyright The OpenTelemetry Authors
// SPDX-License-Identifier: Apache-2.0

// Package genaiconv provides types and functionality for OpenTelemetry semantic
// conventions in the "gen_ai" namespace.
package genaiconv

import (
	"context"

	"go.opentelemetry.io/otel/attribute"
	"go.opentelemetry.io/otel/metric"
	"go.opentelemetry.io/otel/metric/noop"
	"go.opentelemetry.io/otel/semconv/internal/metricpool"
)

// ErrorTypeAttr is an attribute conforming to the error.type semantic
// conventions. It represents the describes a class of error the operation ended
// with.
type ErrorTypeAttr string

var (
	// ErrorTypeOther is a fallback error value to be used when the instrumentation
	// doesn't define a custom value.
	ErrorTypeOther ErrorTypeAttr = "_OTHER"
)

// OperationNameAttr is an attribute conforming to the gen_ai.operation.name
// semantic conventions. It represents the name of the operation being performed.
type OperationNameAttr string

var (
	// OperationNameChat is the chat completion operation such as [OpenAI Chat API]
	// .
	//
	// [OpenAI Chat API]: https://platform.openai.com/docs/api-reference/chat
	OperationNameChat OperationNameAttr = "chat"
	// OperationNameGenerateContent is the multimodal content generation operation
	// such as [Gemini Generate Content].
	//
	// [Gemini Generate Content]: https://ai.google.dev/api/generate-content
	OperationNameGenerateContent OperationNameAttr = "generate_content"
	// OperationNameTextCompletion is the text completions operation such as
	// [OpenAI Completions API (Legacy)].
	//
	// [OpenAI Completions API (Legacy)]: https://platform.openai.com/docs/api-reference/completions
	OperationNameTextCompletion OperationNameAttr = "text_completion"
	// OperationNameEmbeddings is the embeddings operation such as
	// [OpenAI Create embeddings API].
	//
	// [OpenAI Create embeddings API]: https://platform.openai.com/docs/api-reference/embeddings/create
	OperationNameEmbeddings OperationNameAttr = "embeddings"
	// OperationNameRetrieval is the retrieval operation such as
	// [OpenAI Search Vector Store API].
	//
	// [OpenAI Search Vector Store API]: https://platform.openai.com/docs/api-reference/vector-stores/search
	OperationNameRetrieval OperationNameAttr = "retrieval"
	// OperationNameCreateAgent is the create GenAI agent.
	OperationNameCreateAgent OperationNameAttr = "create_agent"
	// OperationNameInvokeAgent is the invoke GenAI agent.
	OperationNameInvokeAgent OperationNameAttr = "invoke_agent"
	// OperationNameExecuteTool is the execute a tool.
	OperationNameExecuteTool OperationNameAttr = "execute_tool"
	// OperationNameInvokeWorkflow is the invoke GenAI workflow.
	OperationNameInvokeWorkflow OperationNameAttr = "invoke_workflow"
)

// ProviderNameAttr is an attribute conforming to the gen_ai.provider.name
// semantic conventions. It represents the Generative AI provider as identified
// by the client or server instrumentation.
type ProviderNameAttr string

var (
	// ProviderNameOpenAI is the [OpenAI].
	//
	// [OpenAI]: https://openai.com/
	ProviderNameOpenAI ProviderNameAttr = "openai"
	// ProviderNameGCPGenAI is the any Google generative AI endpoint.
	ProviderNameGCPGenAI ProviderNameAttr = "gcp.gen_ai"
	// ProviderNameGCPVertexAI is the [Vertex AI].
	//
	// [Vertex AI]: https://cloud.google.com/vertex-ai
	ProviderNameGCPVertexAI ProviderNameAttr = "gcp.vertex_ai"
	// ProviderNameGCPGemini is the [Gemini].
	//
	// [Gemini]: https://cloud.google.com/products/gemini
	ProviderNameGCPGemini ProviderNameAttr = "gcp.gemini"
	// ProviderNameAnthropic is the [Anthropic].
	//
	// [Anthropic]: https://www.anthropic.com/
	ProviderNameAnthropic ProviderNameAttr = "anthropic"
	// ProviderNameCohere is the [Cohere].
	//
	// [Cohere]: https://cohere.com/
	ProviderNameCohere ProviderNameAttr = "cohere"
	// ProviderNameAzureAIInference is the azure AI Inference.
	ProviderNameAzureAIInference ProviderNameAttr = "azure.ai.inference"
	// ProviderNameAzureAIOpenAI is the [Azure OpenAI].
	//
	// [Azure OpenAI]: https://learn.microsoft.com/en-us/azure/ai-services/openai/overview
	ProviderNameAzureAIOpenAI ProviderNameAttr = "azure.ai.openai"
	// ProviderNameIBMWatsonxAI is the [IBM Watsonx AI].
	//
	// [IBM Watsonx AI]: https://www.ibm.com/products/watsonx-ai
	ProviderNameIBMWatsonxAI ProviderNameAttr = "ibm.watsonx.ai"
	// ProviderNameAWSBedrock is the [AWS Bedrock].
	//
	// [AWS Bedrock]: https://aws.amazon.com/bedrock
	ProviderNameAWSBedrock ProviderNameAttr = "aws.bedrock"
	// ProviderNamePerplexity is the [Perplexity].
	//
	// [Perplexity]: https://www.perplexity.ai/
	ProviderNamePerplexity ProviderNameAttr = "perplexity"
	// ProviderNameXAI is the [xAI].
	//
	// [xAI]: https://x.ai/
	ProviderNameXAI ProviderNameAttr = "x_ai"
	// ProviderNameDeepseek is the [DeepSeek].
	//
	// [DeepSeek]: https://www.deepseek.com/
	ProviderNameDeepseek ProviderNameAttr = "deepseek"
	// ProviderNameGroq is the [Groq].
	//
	// [Groq]: https://groq.com/
	ProviderNameGroq ProviderNameAttr = "groq"
	// ProviderNameMistralAI is the [Mistral AI].
	//
	// [Mistral AI]: https://mistral.ai/
	ProviderNameMistralAI ProviderNameAttr = "mistral_ai"
)

// TokenTypeAttr is an attribute conforming to the gen_ai.token.type semantic
// conventions. It represents the type of token being counted.
type TokenTypeAttr string

var (
	// TokenTypeInput is the input tokens (prompt, input, etc.).
	TokenTypeInput TokenTypeAttr = "input"
	// TokenTypeOutput is the output tokens (completion, response, etc.).
	TokenTypeOutput TokenTypeAttr = "output"
)

// ClientOperationDuration is an instrument used to record metric values
// conforming to the "gen_ai.client.operation.duration" semantic conventions. It
// represents the genAI operation duration.
type ClientOperationDuration struct {
	metric.Float64Histogram
}

var newClientOperationDurationOpts = []metric.Float64HistogramOption{
	metric.WithDescription("GenAI operation duration."),
	metric.WithUnit("s"),
}

// NewClientOperationDuration returns a new ClientOperationDuration instrument.
func NewClientOperationDuration(
	m metric.Meter,
	opt ...metric.Float64HistogramOption,
) (ClientOperationDuration, error) {
	// Check if the meter is nil.
	if m == nil {
		return ClientOperationDuration{noop.Float64Histogram{}}, nil
	}

	if len(opt) == 0 {
		opt = newClientOperationDurationOpts
	} else {
		opt = append(opt, newClientOperationDurationOpts...)
	}

	i, err := m.Float64Histogram(
		"gen_ai.client.operation.duration",
		opt...,
	)
	if err != nil {
		return ClientOperationDuration{noop.Float64Histogram{}}, err
	}
	return ClientOperationDuration{i}, nil
}

// Inst returns the underlying metric instrument.
func (m ClientOperationDuration) Inst() metric.Float64Histogram {
	return m.Float64Histogram
}

// Name returns the semantic convention name of the instrument.
func (ClientOperationDuration) Name() string {
	return "gen_ai.client.operation.duration"
}

// Unit returns the semantic convention unit of the instrument
func (ClientOperationDuration) Unit() string {
	return "s"
}

// Description returns the semantic convention description of the instrument
func (ClientOperationDuration) Description() string {
	return "GenAI operation duration."
}

// Record records val to the current distribution for attrs.
//
// The operationName is the the name of the operation being performed.
//
// The providerName is the the Generative AI provider as identified by the client
// or server instrumentation.
//
// All additional attrs passed are included in the recorded value.
func (m ClientOperationDuration) Record(
	ctx context.Context,
	val float64,
	operationName OperationNameAttr,
	providerName ProviderNameAttr,
	attrs ...attribute.KeyValue,
) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if len(attrs) == 0 {
		m.Float64Histogram.Record(ctx, val, metric.WithAttributes(
			attribute.String("gen_ai.operation.name", string(operationName)),
			attribute.String("gen_ai.provider.name", string(providerName)),
		))
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(
		*o,
		metric.WithAttributes(
			append(
				attrs[:len(attrs):len(attrs)],
				attribute.String("gen_ai.operation.name", string(operationName)),
				attribute.String("gen_ai.provider.name", string(providerName)),
			)...,
		),
	)

	m.Float64Histogram.Record(ctx, val, *o...)
}

// RecordSet records val to the current distribution for set.
func (m ClientOperationDuration) RecordSet(ctx context.Context, val float64, set attribute.Set) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if set.Len() == 0 {
		m.Float64Histogram.Record(ctx, val)
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(*o, metric.WithAttributeSet(set))
	m.Float64Histogram.Record(ctx, val, *o...)
}

// AttrErrorType returns an optional attribute for the "error.type" semantic
// convention. It represents the describes a class of error the operation ended
// with.
func (ClientOperationDuration) AttrErrorType(val ErrorTypeAttr) attribute.KeyValue {
	return attribute.String("error.type", string(val))
}

// AttrRequestModel returns an optional attribute for the "gen_ai.request.model"
// semantic convention. It represents the name of the GenAI model a request is
// being made to.
func (ClientOperationDuration) AttrRequestModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.request.model", val)
}

// AttrServerPort returns an optional attribute for the "server.port" semantic
// convention. It represents the genAI server port.
func (ClientOperationDuration) AttrServerPort(val int) attribute.KeyValue {
	return attribute.Int("server.port", val)
}

// AttrResponseModel returns an optional attribute for the
// "gen_ai.response.model" semantic convention. It represents the name of the
// model that generated the response.
func (ClientOperationDuration) AttrResponseModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.response.model", val)
}

// AttrServerAddress returns an optional attribute for the "server.address"
// semantic convention. It represents the genAI server address.
func (ClientOperationDuration) AttrServerAddress(val string) attribute.KeyValue {
	return attribute.String("server.address", val)
}

// ClientOperationTimePerOutputChunk is an instrument used to record metric
// values conforming to the "gen_ai.client.operation.time_per_output_chunk"
// semantic conventions. It represents the time 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 current chunk.
type ClientOperationTimePerOutputChunk struct {
	metric.Float64Histogram
}

var newClientOperationTimePerOutputChunkOpts = []metric.Float64HistogramOption{
	metric.WithDescription("Time 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 current chunk."),
	metric.WithUnit("s"),
}

// NewClientOperationTimePerOutputChunk returns a new
// ClientOperationTimePerOutputChunk instrument.
func NewClientOperationTimePerOutputChunk(
	m metric.Meter,
	opt ...metric.Float64HistogramOption,
) (ClientOperationTimePerOutputChunk, error) {
	// Check if the meter is nil.
	if m == nil {
		return ClientOperationTimePerOutputChunk{noop.Float64Histogram{}}, nil
	}

	if len(opt) == 0 {
		opt = newClientOperationTimePerOutputChunkOpts
	} else {
		opt = append(opt, newClientOperationTimePerOutputChunkOpts...)
	}

	i, err := m.Float64Histogram(
		"gen_ai.client.operation.time_per_output_chunk",
		opt...,
	)
	if err != nil {
		return ClientOperationTimePerOutputChunk{noop.Float64Histogram{}}, err
	}
	return ClientOperationTimePerOutputChunk{i}, nil
}

// Inst returns the underlying metric instrument.
func (m ClientOperationTimePerOutputChunk) Inst() metric.Float64Histogram {
	return m.Float64Histogram
}

// Name returns the semantic convention name of the instrument.
func (ClientOperationTimePerOutputChunk) Name() string {
	return "gen_ai.client.operation.time_per_output_chunk"
}

// Unit returns the semantic convention unit of the instrument
func (ClientOperationTimePerOutputChunk) Unit() string {
	return "s"
}

// Description returns the semantic convention description of the instrument
func (ClientOperationTimePerOutputChunk) Description() string {
	return "Time 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 current chunk."
}

// Record records val to the current distribution for attrs.
//
// The operationName is the the name of the operation being performed.
//
// The providerName is the the Generative AI provider as identified by the client
// or server instrumentation.
//
// All additional attrs passed are included in the recorded value.
//
// This metrics SHOULD be reported for streaming calls and SHOULD NOT be reported
// otherwise.
func (m ClientOperationTimePerOutputChunk) Record(
	ctx context.Context,
	val float64,
	operationName OperationNameAttr,
	providerName ProviderNameAttr,
	attrs ...attribute.KeyValue,
) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if len(attrs) == 0 {
		m.Float64Histogram.Record(ctx, val, metric.WithAttributes(
			attribute.String("gen_ai.operation.name", string(operationName)),
			attribute.String("gen_ai.provider.name", string(providerName)),
		))
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(
		*o,
		metric.WithAttributes(
			append(
				attrs[:len(attrs):len(attrs)],
				attribute.String("gen_ai.operation.name", string(operationName)),
				attribute.String("gen_ai.provider.name", string(providerName)),
			)...,
		),
	)

	m.Float64Histogram.Record(ctx, val, *o...)
}

// RecordSet records val to the current distribution for set.
//
// This metrics SHOULD be reported for streaming calls and SHOULD NOT be reported
// otherwise.
func (m ClientOperationTimePerOutputChunk) RecordSet(ctx context.Context, val float64, set attribute.Set) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if set.Len() == 0 {
		m.Float64Histogram.Record(ctx, val)
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(*o, metric.WithAttributeSet(set))
	m.Float64Histogram.Record(ctx, val, *o...)
}

// AttrRequestModel returns an optional attribute for the "gen_ai.request.model"
// semantic convention. It represents the name of the GenAI model a request is
// being made to.
func (ClientOperationTimePerOutputChunk) AttrRequestModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.request.model", val)
}

// AttrServerPort returns an optional attribute for the "server.port" semantic
// convention. It represents the genAI server port.
func (ClientOperationTimePerOutputChunk) AttrServerPort(val int) attribute.KeyValue {
	return attribute.Int("server.port", val)
}

// AttrResponseModel returns an optional attribute for the
// "gen_ai.response.model" semantic convention. It represents the name of the
// model that generated the response.
func (ClientOperationTimePerOutputChunk) AttrResponseModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.response.model", val)
}

// AttrServerAddress returns an optional attribute for the "server.address"
// semantic convention. It represents the genAI server address.
func (ClientOperationTimePerOutputChunk) AttrServerAddress(val string) attribute.KeyValue {
	return attribute.String("server.address", val)
}

// ClientOperationTimeToFirstChunk is an instrument used to record metric values
// conforming to the "gen_ai.client.operation.time_to_first_chunk" semantic
// conventions. It represents the time 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.
type ClientOperationTimeToFirstChunk struct {
	metric.Float64Histogram
}

var newClientOperationTimeToFirstChunkOpts = []metric.Float64HistogramOption{
	metric.WithDescription("Time 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."),
	metric.WithUnit("s"),
}

// NewClientOperationTimeToFirstChunk returns a new
// ClientOperationTimeToFirstChunk instrument.
func NewClientOperationTimeToFirstChunk(
	m metric.Meter,
	opt ...metric.Float64HistogramOption,
) (ClientOperationTimeToFirstChunk, error) {
	// Check if the meter is nil.
	if m == nil {
		return ClientOperationTimeToFirstChunk{noop.Float64Histogram{}}, nil
	}

	if len(opt) == 0 {
		opt = newClientOperationTimeToFirstChunkOpts
	} else {
		opt = append(opt, newClientOperationTimeToFirstChunkOpts...)
	}

	i, err := m.Float64Histogram(
		"gen_ai.client.operation.time_to_first_chunk",
		opt...,
	)
	if err != nil {
		return ClientOperationTimeToFirstChunk{noop.Float64Histogram{}}, err
	}
	return ClientOperationTimeToFirstChunk{i}, nil
}

// Inst returns the underlying metric instrument.
func (m ClientOperationTimeToFirstChunk) Inst() metric.Float64Histogram {
	return m.Float64Histogram
}

// Name returns the semantic convention name of the instrument.
func (ClientOperationTimeToFirstChunk) Name() string {
	return "gen_ai.client.operation.time_to_first_chunk"
}

// Unit returns the semantic convention unit of the instrument
func (ClientOperationTimeToFirstChunk) Unit() string {
	return "s"
}

// Description returns the semantic convention description of the instrument
func (ClientOperationTimeToFirstChunk) Description() string {
	return "Time 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."
}

// Record records val to the current distribution for attrs.
//
// The operationName is the the name of the operation being performed.
//
// The providerName is the the Generative AI provider as identified by the client
// or server instrumentation.
//
// All additional attrs passed are included in the recorded value.
//
// This metrics SHOULD be reported for streaming calls and SHOULD NOT be reported
// otherwise.
func (m ClientOperationTimeToFirstChunk) Record(
	ctx context.Context,
	val float64,
	operationName OperationNameAttr,
	providerName ProviderNameAttr,
	attrs ...attribute.KeyValue,
) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if len(attrs) == 0 {
		m.Float64Histogram.Record(ctx, val, metric.WithAttributes(
			attribute.String("gen_ai.operation.name", string(operationName)),
			attribute.String("gen_ai.provider.name", string(providerName)),
		))
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(
		*o,
		metric.WithAttributes(
			append(
				attrs[:len(attrs):len(attrs)],
				attribute.String("gen_ai.operation.name", string(operationName)),
				attribute.String("gen_ai.provider.name", string(providerName)),
			)...,
		),
	)

	m.Float64Histogram.Record(ctx, val, *o...)
}

// RecordSet records val to the current distribution for set.
//
// This metrics SHOULD be reported for streaming calls and SHOULD NOT be reported
// otherwise.
func (m ClientOperationTimeToFirstChunk) RecordSet(ctx context.Context, val float64, set attribute.Set) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if set.Len() == 0 {
		m.Float64Histogram.Record(ctx, val)
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(*o, metric.WithAttributeSet(set))
	m.Float64Histogram.Record(ctx, val, *o...)
}

// AttrRequestModel returns an optional attribute for the "gen_ai.request.model"
// semantic convention. It represents the name of the GenAI model a request is
// being made to.
func (ClientOperationTimeToFirstChunk) AttrRequestModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.request.model", val)
}

// AttrServerPort returns an optional attribute for the "server.port" semantic
// convention. It represents the genAI server port.
func (ClientOperationTimeToFirstChunk) AttrServerPort(val int) attribute.KeyValue {
	return attribute.Int("server.port", val)
}

// AttrResponseModel returns an optional attribute for the
// "gen_ai.response.model" semantic convention. It represents the name of the
// model that generated the response.
func (ClientOperationTimeToFirstChunk) AttrResponseModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.response.model", val)
}

// AttrServerAddress returns an optional attribute for the "server.address"
// semantic convention. It represents the genAI server address.
func (ClientOperationTimeToFirstChunk) AttrServerAddress(val string) attribute.KeyValue {
	return attribute.String("server.address", val)
}

// ClientTokenUsage is an instrument used to record metric values conforming to
// the "gen_ai.client.token.usage" semantic conventions. It represents the number
// of input and output tokens used.
type ClientTokenUsage struct {
	metric.Int64Histogram
}

var newClientTokenUsageOpts = []metric.Int64HistogramOption{
	metric.WithDescription("Number of input and output tokens used."),
	metric.WithUnit("{token}"),
}

// NewClientTokenUsage returns a new ClientTokenUsage instrument.
func NewClientTokenUsage(
	m metric.Meter,
	opt ...metric.Int64HistogramOption,
) (ClientTokenUsage, error) {
	// Check if the meter is nil.
	if m == nil {
		return ClientTokenUsage{noop.Int64Histogram{}}, nil
	}

	if len(opt) == 0 {
		opt = newClientTokenUsageOpts
	} else {
		opt = append(opt, newClientTokenUsageOpts...)
	}

	i, err := m.Int64Histogram(
		"gen_ai.client.token.usage",
		opt...,
	)
	if err != nil {
		return ClientTokenUsage{noop.Int64Histogram{}}, err
	}
	return ClientTokenUsage{i}, nil
}

// Inst returns the underlying metric instrument.
func (m ClientTokenUsage) Inst() metric.Int64Histogram {
	return m.Int64Histogram
}

// Name returns the semantic convention name of the instrument.
func (ClientTokenUsage) Name() string {
	return "gen_ai.client.token.usage"
}

// Unit returns the semantic convention unit of the instrument
func (ClientTokenUsage) Unit() string {
	return "{token}"
}

// Description returns the semantic convention description of the instrument
func (ClientTokenUsage) Description() string {
	return "Number of input and output tokens used."
}

// Record records val to the current distribution for attrs.
//
// The operationName is the the name of the operation being performed.
//
// The providerName is the the Generative AI provider as identified by the client
// or server instrumentation.
//
// The tokenType is the the type of token being counted.
//
// All additional attrs passed are included in the recorded value.
func (m ClientTokenUsage) Record(
	ctx context.Context,
	val int64,
	operationName OperationNameAttr,
	providerName ProviderNameAttr,
	tokenType TokenTypeAttr,
	attrs ...attribute.KeyValue,
) {
	if !m.Int64Histogram.Enabled(ctx) {
		return
	}
	if len(attrs) == 0 {
		m.Int64Histogram.Record(ctx, val, metric.WithAttributes(
			attribute.String("gen_ai.operation.name", string(operationName)),
			attribute.String("gen_ai.provider.name", string(providerName)),
			attribute.String("gen_ai.token.type", string(tokenType)),
		))
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(
		*o,
		metric.WithAttributes(
			append(
				attrs[:len(attrs):len(attrs)],
				attribute.String("gen_ai.operation.name", string(operationName)),
				attribute.String("gen_ai.provider.name", string(providerName)),
				attribute.String("gen_ai.token.type", string(tokenType)),
			)...,
		),
	)

	m.Int64Histogram.Record(ctx, val, *o...)
}

// RecordSet records val to the current distribution for set.
func (m ClientTokenUsage) RecordSet(ctx context.Context, val int64, set attribute.Set) {
	if !m.Int64Histogram.Enabled(ctx) {
		return
	}
	if set.Len() == 0 {
		m.Int64Histogram.Record(ctx, val)
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(*o, metric.WithAttributeSet(set))
	m.Int64Histogram.Record(ctx, val, *o...)
}

// AttrRequestModel returns an optional attribute for the "gen_ai.request.model"
// semantic convention. It represents the name of the GenAI model a request is
// being made to.
func (ClientTokenUsage) AttrRequestModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.request.model", val)
}

// AttrServerPort returns an optional attribute for the "server.port" semantic
// convention. It represents the genAI server port.
func (ClientTokenUsage) AttrServerPort(val int) attribute.KeyValue {
	return attribute.Int("server.port", val)
}

// AttrResponseModel returns an optional attribute for the
// "gen_ai.response.model" semantic convention. It represents the name of the
// model that generated the response.
func (ClientTokenUsage) AttrResponseModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.response.model", val)
}

// AttrServerAddress returns an optional attribute for the "server.address"
// semantic convention. It represents the genAI server address.
func (ClientTokenUsage) AttrServerAddress(val string) attribute.KeyValue {
	return attribute.String("server.address", val)
}

// ServerRequestDuration is an instrument used to record metric values conforming
// to the "gen_ai.server.request.duration" semantic conventions. It represents
// the generative AI server request duration such as time-to-last byte or last
// output token.
type ServerRequestDuration struct {
	metric.Float64Histogram
}

var newServerRequestDurationOpts = []metric.Float64HistogramOption{
	metric.WithDescription("Generative AI server request duration such as time-to-last byte or last output token."),
	metric.WithUnit("s"),
}

// NewServerRequestDuration returns a new ServerRequestDuration instrument.
func NewServerRequestDuration(
	m metric.Meter,
	opt ...metric.Float64HistogramOption,
) (ServerRequestDuration, error) {
	// Check if the meter is nil.
	if m == nil {
		return ServerRequestDuration{noop.Float64Histogram{}}, nil
	}

	if len(opt) == 0 {
		opt = newServerRequestDurationOpts
	} else {
		opt = append(opt, newServerRequestDurationOpts...)
	}

	i, err := m.Float64Histogram(
		"gen_ai.server.request.duration",
		opt...,
	)
	if err != nil {
		return ServerRequestDuration{noop.Float64Histogram{}}, err
	}
	return ServerRequestDuration{i}, nil
}

// Inst returns the underlying metric instrument.
func (m ServerRequestDuration) Inst() metric.Float64Histogram {
	return m.Float64Histogram
}

// Name returns the semantic convention name of the instrument.
func (ServerRequestDuration) Name() string {
	return "gen_ai.server.request.duration"
}

// Unit returns the semantic convention unit of the instrument
func (ServerRequestDuration) Unit() string {
	return "s"
}

// Description returns the semantic convention description of the instrument
func (ServerRequestDuration) Description() string {
	return "Generative AI server request duration such as time-to-last byte or last output token."
}

// Record records val to the current distribution for attrs.
//
// The operationName is the the name of the operation being performed.
//
// The providerName is the the Generative AI provider as identified by the client
// or server instrumentation.
//
// All additional attrs passed are included in the recorded value.
func (m ServerRequestDuration) Record(
	ctx context.Context,
	val float64,
	operationName OperationNameAttr,
	providerName ProviderNameAttr,
	attrs ...attribute.KeyValue,
) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if len(attrs) == 0 {
		m.Float64Histogram.Record(ctx, val, metric.WithAttributes(
			attribute.String("gen_ai.operation.name", string(operationName)),
			attribute.String("gen_ai.provider.name", string(providerName)),
		))
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(
		*o,
		metric.WithAttributes(
			append(
				attrs[:len(attrs):len(attrs)],
				attribute.String("gen_ai.operation.name", string(operationName)),
				attribute.String("gen_ai.provider.name", string(providerName)),
			)...,
		),
	)

	m.Float64Histogram.Record(ctx, val, *o...)
}

// RecordSet records val to the current distribution for set.
func (m ServerRequestDuration) RecordSet(ctx context.Context, val float64, set attribute.Set) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if set.Len() == 0 {
		m.Float64Histogram.Record(ctx, val)
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(*o, metric.WithAttributeSet(set))
	m.Float64Histogram.Record(ctx, val, *o...)
}

// AttrErrorType returns an optional attribute for the "error.type" semantic
// convention. It represents the describes a class of error the operation ended
// with.
func (ServerRequestDuration) AttrErrorType(val ErrorTypeAttr) attribute.KeyValue {
	return attribute.String("error.type", string(val))
}

// AttrRequestModel returns an optional attribute for the "gen_ai.request.model"
// semantic convention. It represents the name of the GenAI model a request is
// being made to.
func (ServerRequestDuration) AttrRequestModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.request.model", val)
}

// AttrServerPort returns an optional attribute for the "server.port" semantic
// convention. It represents the genAI server port.
func (ServerRequestDuration) AttrServerPort(val int) attribute.KeyValue {
	return attribute.Int("server.port", val)
}

// AttrResponseModel returns an optional attribute for the
// "gen_ai.response.model" semantic convention. It represents the name of the
// model that generated the response.
func (ServerRequestDuration) AttrResponseModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.response.model", val)
}

// AttrServerAddress returns an optional attribute for the "server.address"
// semantic convention. It represents the genAI server address.
func (ServerRequestDuration) AttrServerAddress(val string) attribute.KeyValue {
	return attribute.String("server.address", val)
}

// ServerTimePerOutputToken is an instrument used to record metric values
// conforming to the "gen_ai.server.time_per_output_token" semantic conventions.
// It represents the time per output token generated after the first token for
// successful responses.
type ServerTimePerOutputToken struct {
	metric.Float64Histogram
}

var newServerTimePerOutputTokenOpts = []metric.Float64HistogramOption{
	metric.WithDescription("Time per output token generated after the first token for successful responses."),
	metric.WithUnit("s"),
}

// NewServerTimePerOutputToken returns a new ServerTimePerOutputToken instrument.
func NewServerTimePerOutputToken(
	m metric.Meter,
	opt ...metric.Float64HistogramOption,
) (ServerTimePerOutputToken, error) {
	// Check if the meter is nil.
	if m == nil {
		return ServerTimePerOutputToken{noop.Float64Histogram{}}, nil
	}

	if len(opt) == 0 {
		opt = newServerTimePerOutputTokenOpts
	} else {
		opt = append(opt, newServerTimePerOutputTokenOpts...)
	}

	i, err := m.Float64Histogram(
		"gen_ai.server.time_per_output_token",
		opt...,
	)
	if err != nil {
		return ServerTimePerOutputToken{noop.Float64Histogram{}}, err
	}
	return ServerTimePerOutputToken{i}, nil
}

// Inst returns the underlying metric instrument.
func (m ServerTimePerOutputToken) Inst() metric.Float64Histogram {
	return m.Float64Histogram
}

// Name returns the semantic convention name of the instrument.
func (ServerTimePerOutputToken) Name() string {
	return "gen_ai.server.time_per_output_token"
}

// Unit returns the semantic convention unit of the instrument
func (ServerTimePerOutputToken) Unit() string {
	return "s"
}

// Description returns the semantic convention description of the instrument
func (ServerTimePerOutputToken) Description() string {
	return "Time per output token generated after the first token for successful responses."
}

// Record records val to the current distribution for attrs.
//
// The operationName is the the name of the operation being performed.
//
// The providerName is the the Generative AI provider as identified by the client
// or server instrumentation.
//
// All additional attrs passed are included in the recorded value.
func (m ServerTimePerOutputToken) Record(
	ctx context.Context,
	val float64,
	operationName OperationNameAttr,
	providerName ProviderNameAttr,
	attrs ...attribute.KeyValue,
) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if len(attrs) == 0 {
		m.Float64Histogram.Record(ctx, val, metric.WithAttributes(
			attribute.String("gen_ai.operation.name", string(operationName)),
			attribute.String("gen_ai.provider.name", string(providerName)),
		))
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(
		*o,
		metric.WithAttributes(
			append(
				attrs[:len(attrs):len(attrs)],
				attribute.String("gen_ai.operation.name", string(operationName)),
				attribute.String("gen_ai.provider.name", string(providerName)),
			)...,
		),
	)

	m.Float64Histogram.Record(ctx, val, *o...)
}

// RecordSet records val to the current distribution for set.
func (m ServerTimePerOutputToken) RecordSet(ctx context.Context, val float64, set attribute.Set) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if set.Len() == 0 {
		m.Float64Histogram.Record(ctx, val)
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(*o, metric.WithAttributeSet(set))
	m.Float64Histogram.Record(ctx, val, *o...)
}

// AttrRequestModel returns an optional attribute for the "gen_ai.request.model"
// semantic convention. It represents the name of the GenAI model a request is
// being made to.
func (ServerTimePerOutputToken) AttrRequestModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.request.model", val)
}

// AttrServerPort returns an optional attribute for the "server.port" semantic
// convention. It represents the genAI server port.
func (ServerTimePerOutputToken) AttrServerPort(val int) attribute.KeyValue {
	return attribute.Int("server.port", val)
}

// AttrResponseModel returns an optional attribute for the
// "gen_ai.response.model" semantic convention. It represents the name of the
// model that generated the response.
func (ServerTimePerOutputToken) AttrResponseModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.response.model", val)
}

// AttrServerAddress returns an optional attribute for the "server.address"
// semantic convention. It represents the genAI server address.
func (ServerTimePerOutputToken) AttrServerAddress(val string) attribute.KeyValue {
	return attribute.String("server.address", val)
}

// ServerTimeToFirstToken is an instrument used to record metric values
// conforming to the "gen_ai.server.time_to_first_token" semantic conventions. It
// represents the time to generate first token for successful responses.
type ServerTimeToFirstToken struct {
	metric.Float64Histogram
}

var newServerTimeToFirstTokenOpts = []metric.Float64HistogramOption{
	metric.WithDescription("Time to generate first token for successful responses."),
	metric.WithUnit("s"),
}

// NewServerTimeToFirstToken returns a new ServerTimeToFirstToken instrument.
func NewServerTimeToFirstToken(
	m metric.Meter,
	opt ...metric.Float64HistogramOption,
) (ServerTimeToFirstToken, error) {
	// Check if the meter is nil.
	if m == nil {
		return ServerTimeToFirstToken{noop.Float64Histogram{}}, nil
	}

	if len(opt) == 0 {
		opt = newServerTimeToFirstTokenOpts
	} else {
		opt = append(opt, newServerTimeToFirstTokenOpts...)
	}

	i, err := m.Float64Histogram(
		"gen_ai.server.time_to_first_token",
		opt...,
	)
	if err != nil {
		return ServerTimeToFirstToken{noop.Float64Histogram{}}, err
	}
	return ServerTimeToFirstToken{i}, nil
}

// Inst returns the underlying metric instrument.
func (m ServerTimeToFirstToken) Inst() metric.Float64Histogram {
	return m.Float64Histogram
}

// Name returns the semantic convention name of the instrument.
func (ServerTimeToFirstToken) Name() string {
	return "gen_ai.server.time_to_first_token"
}

// Unit returns the semantic convention unit of the instrument
func (ServerTimeToFirstToken) Unit() string {
	return "s"
}

// Description returns the semantic convention description of the instrument
func (ServerTimeToFirstToken) Description() string {
	return "Time to generate first token for successful responses."
}

// Record records val to the current distribution for attrs.
//
// The operationName is the the name of the operation being performed.
//
// The providerName is the the Generative AI provider as identified by the client
// or server instrumentation.
//
// All additional attrs passed are included in the recorded value.
func (m ServerTimeToFirstToken) Record(
	ctx context.Context,
	val float64,
	operationName OperationNameAttr,
	providerName ProviderNameAttr,
	attrs ...attribute.KeyValue,
) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if len(attrs) == 0 {
		m.Float64Histogram.Record(ctx, val, metric.WithAttributes(
			attribute.String("gen_ai.operation.name", string(operationName)),
			attribute.String("gen_ai.provider.name", string(providerName)),
		))
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(
		*o,
		metric.WithAttributes(
			append(
				attrs[:len(attrs):len(attrs)],
				attribute.String("gen_ai.operation.name", string(operationName)),
				attribute.String("gen_ai.provider.name", string(providerName)),
			)...,
		),
	)

	m.Float64Histogram.Record(ctx, val, *o...)
}

// RecordSet records val to the current distribution for set.
func (m ServerTimeToFirstToken) RecordSet(ctx context.Context, val float64, set attribute.Set) {
	if !m.Float64Histogram.Enabled(ctx) {
		return
	}
	if set.Len() == 0 {
		m.Float64Histogram.Record(ctx, val)
		return
	}

	o := metricpool.RecordOptions()
	defer metricpool.PutRecordOptions(o)

	*o = append(*o, metric.WithAttributeSet(set))
	m.Float64Histogram.Record(ctx, val, *o...)
}

// AttrRequestModel returns an optional attribute for the "gen_ai.request.model"
// semantic convention. It represents the name of the GenAI model a request is
// being made to.
func (ServerTimeToFirstToken) AttrRequestModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.request.model", val)
}

// AttrServerPort returns an optional attribute for the "server.port" semantic
// convention. It represents the genAI server port.
func (ServerTimeToFirstToken) AttrServerPort(val int) attribute.KeyValue {
	return attribute.Int("server.port", val)
}

// AttrResponseModel returns an optional attribute for the
// "gen_ai.response.model" semantic convention. It represents the name of the
// model that generated the response.
func (ServerTimeToFirstToken) AttrResponseModel(val string) attribute.KeyValue {
	return attribute.String("gen_ai.response.model", val)
}

// AttrServerAddress returns an optional attribute for the "server.address"
// semantic convention. It represents the genAI server address.
func (ServerTimeToFirstToken) AttrServerAddress(val string) attribute.KeyValue {
	return attribute.String("server.address", val)
}