打开推论 GenAI

这个软件包提供一套转换的公用设施 OpenTeleometry GenAI 函数 横跨属性到 OpenInference 横跨属性.

[尖叫声] OpenTeleometry GenAI公约仍在孵化中,可能包括随时突破变化. 此软件包将尝试将 OpenTeleometry GenAI 属性的一个子集转换为 OpenInference 属性的最佳努力. 目前,属性反映了截至2025年10月的定义.

安装

npm install --save @arizeai/openinference-genai

使用量

@arizeai/openinference-genai 可以作为一组独立的帮助器函数, 或与 SpanProcessor 相配合,以便自动将 OpenTeleometry GenAI 跨度转换为 OpenInference 跨度.

独立

您可以通过使用独立的帮助器函数来突变设置的跨度属性 .

[大 Span 突变不为 OpenTelemetery SDK 所支持, 因此确保您是 在跨度寿命的最后一英里进行突变(即在将跨度导出为 SpanProcessor 之前).

import { convertGenAISpanAttributesToOpenInferenceSpanAttributes } from `@arizeai/openinference-genai`

// obtain a span with OpenTelemetry GenAI attributes from your tracing system
const span: ReadableSpan = {/* ... */}

// convert the span attributes to OpenInference attributes
const openinferenceAttributes = convertGenAISpanAttributesToOpenInferenceSpanAttributes(span.attributes)

// add the OpenInference attributes to the span
span.attributes = {...span.attributes, ...openinferenceAttributes}

块处理器

您可以使用自定义的 TraceExporter 将 OpenTelemetry GenAI 跨度自动转换为 OpenInference 跨度 .

见 实例/导出- spans.ts 用于以下样本代码的可运行版本.

通过安装软件包开始

pnpm add @opentelemetry/api @opentelemetry/core @opentelemetry/exporter-trace-otlp-proto @opentelemetry/sdk-trace-base @opentelemetry/sdk-trace-node @opentelemetry/semantic-conventions @opentelemetry/resources @arizeai/openinference-genai

创建自定义 TraceExporter,将 OpenTelemetry GenAI 属性转换为 OpenInference 属性.

// openinferenceOTLPTraceExporter.ts
import { OTLPTraceExporter } from "@opentelemetry/exporter-trace-otlp-proto";
import type { ReadableSpan } from "@opentelemetry/sdk-trace-base";
import type { ExportResult } from "@opentelemetry/core";

import { convertGenAISpanAttributesToOpenInferenceSpanAttributes } from "@arizeai/openinference-genai";
import type { Mutable } from "@arizeai/openinference-genai/types";

class OpenInferenceOTLPTraceExporter extends OTLPTraceExporter {
  export(spans: ReadableSpan[], resultCallback: (result: ExportResult) => void) {
    const processedSpans = spans.map((span) => {
      const processedAttributes = convertGenAISpanAttributesToOpenInferenceSpanAttributes(
        span.attributes,
      );
      // optionally you can replace the entire attributes object with the
      // processed attributes if you want _only_ the OpenInference attributes
      (span as Mutable<ReadableSpan>).attributes = {
        ...span.attributes,
        ...processedAttributes,
      };
      return span;
    });

    super.export(processedSpans, resultCallback);
  }
}

然后用在你选择的处理器里.

// instrumentation.ts
import { resourceFromAttributes } from "@opentelemetry/resources";
import { NodeTracerProvider, BatchSpanProcessor } from "@opentelemetry/sdk-trace-node";
import { ATTR_SERVICE_NAME } from "@opentelemetry/semantic-conventions";

import { SEMRESATTRS_PROJECT_NAME } from "@arizeai/openinference-semantic-conventions";

import { OpenInferenceOTLPTraceExporter } from "./openinferenceOTLPTraceExporter";

const COLLECTOR_ENDPOINT = process.env.COLLECTOR_ENDPOINT;
const SERVICE_NAME = "openinference-genai-app";

export const provider = new NodeTracerProvider({
  resource: resourceFromAttributes({
    [ATTR_SERVICE_NAME]: SERVICE_NAME,
    [SEMRESATTRS_PROJECT_NAME]: SERVICE_NAME,
  }),
  spanProcessors: [
    new BatchSpanProcessor(
      new OpenInferenceOTLPTraceExporter({
        url: `${COLLECTOR_ENDPOINT}/v1/traces`,
      }),
    ),
  ],
});

provider.register();

实例

见 实例 用于此软件包中更多可执行示例的目录.

要执行示例, 请运行以下命令:

cd js/packages/openinference-genai
pnpm install
pnpm -r build
pnpx tsx examples/export-spans.ts