Skip to content

integrations.md

Latest commit

 

History

History
89 lines (64 loc) · 2.7 KB

integrations.md

File metadata and controls

89 lines (64 loc) · 2.7 KB
Raw

Tracing integrations

Auto-instrument popular frameworks so their calls become PromptRails spans. Each integration is a separate import and only assumes you've installed that framework.

LangChain

A callback handler that turns LangChain runs (chains, LLMs, tools, retrievers) into a span tree. It implements LangChain's CallbackHandlerMethods, so pass it via callbacks:

import { Tracer } from "@promptrails/sdk/tracing";
import { PromptRailsCallbackHandler } from "@promptrails/sdk/tracing/integrations/langchain";

const tracer = new Tracer({ apiKey: "pr_..." });
const handler = new PromptRailsCallbackHandler(tracer);

await chain.invoke(
  { question: "What's the weather?" },
  { callbacks: [handler] },
);
await tracer.flush();

The tree is built from LangChain's runId/parentRunId, so it is correct under concurrent runs. Token usage and model are read on handleLLMEnd.

OpenAI

Wrap an OpenAI-compatible client so every chat.completions.create call emits an llm span with model, token usage, latency, and output:

import OpenAI from "openai";
import { Tracer } from "@promptrails/sdk/tracing";
import { traceOpenAI } from "@promptrails/sdk/tracing/integrations/openai";

const tracer = new Tracer({ apiKey: "pr_..." });
const client = traceOpenAI(new OpenAI(), tracer);

await client.chat.completions.create({ model: "gpt-4o", messages: [...] });

The wrapper is duck-typed, so it also works with any API-compatible client.

Anthropic

Wrap an Anthropic client to trace every messages.create call:

import Anthropic from "@anthropic-ai/sdk";
import { traceAnthropic } from "@promptrails/sdk/tracing/integrations/anthropic";

const client = traceAnthropic(new Anthropic(), tracer);
await client.messages.create({ model: "claude-sonnet-4-5", max_tokens: 1024, messages: [...] });

Google GenAI

Wrap a Google GenAI client (the @google/genai SDK) to trace every models.generateContent call:

import { GoogleGenAI } from "@google/genai";
import { traceGoogle } from "@promptrails/sdk/tracing/integrations/google";

const client = traceGoogle(new GoogleGenAI({ apiKey: "..." }), tracer);
await client.models.generateContent({
  model: "gemini-2.0-flash",
  contents: "Hello",
});

OpenTelemetry

Already using OpenTelemetry? Register the PromptRails exporter and your existing spans flow in — gen_ai.* semantic-convention attributes are mapped onto the span model:

import { BatchSpanProcessor } from "@opentelemetry/sdk-trace-base";
import { PromptRailsSpanExporter } from "@promptrails/sdk/tracing/integrations/otel";

provider.addSpanProcessor(
  new BatchSpanProcessor(new PromptRailsSpanExporter({ apiKey: "pr_..." })),
);