Spendline JavaScript SDK

Spendline tracks LLM usage in production with a single wrapper. The SDK captures tokens, model, latency, cost, timestamp, workflow ID, and request metadata without collecting prompt or completion text.

Install

npm install spendline

Quick Start

import OpenAI from "openai"
import { track } from "spendline"

const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY })

const completion = await track(
  () =>
    openai.chat.completions.create({
      model: "gpt-5-mini",
      messages: [{ role: "user", content: "Say hello" }],
    }),
  {
    workflowId: "support-bot",
    sessionId: "session-123",
    metadata: { feature: "chat", environment: "production" },
  },
)

console.log(completion.choices[0]?.message?.content)

Auto-Patch OpenAI

import OpenAI from "openai"
import { patchOpenAI } from "spendline"

const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY })
patchOpenAI(openai)

Use a Dedicated Client

import { Spendline } from "spendline"

const spendline = new Spendline({
  apiKey: process.env.SPENDLINE_API_KEY,
  apiUrl: "https://api.spendline.dev",
})

await spendline.track(() =>
  openai.chat.completions.create({
    model: "gpt-5-mini",
    messages: [{ role: "user", content: "Hello" }],
  }),
)

Supported Providers

• OpenAI via track() and patchOpenAI()
• Anthropic via track() and patchAnthropic()
• Bedrock, Gemini, and DeepSeek via track()

Environment Variables

| Variable | Purpose | | --- | --- | | SPENDLINE_API_KEY | Default API key for ingest | | SPENDLINE_API_URL | Override the API base URL | | SPENDLINE_DISABLE | Disable tracking entirely when set to true | | SPENDLINE_LOG | Print events locally before flush |

Privacy

Spendline does not collect:

• Prompt text
• Completion text
• Raw request bodies
• Application secrets