@uma/toolkit

Shared, framework‑agnostic helpers that we reuse across Risk Lab's TypeScript codebases.

The first module shipping here is a rate‑limited, retry‑friendly HTTP client built on Axios + Bottleneck + axios‑retry—perfect for hitting third‑party APIs without blowing through their quotas.

Installation

# inside any workspace in the monorepo
yarn lerna add @uma/toolkit --scope @uma/monitor-v2

External project? > yarn add @uma/toolkit once it’s published to npm.

Quick start

import { http } from "@uma/toolkit/http"

const { data } = await http.get("https://example.com/api/v1/status")

Defaults

• max 5 concurrent requests
• ≥ 200 ms gap → ≈ 5 req/s
• timeout 10 s
• 3 retries on network errors, 5xx or 429 with exponential back‑off + jitter

Custom client per service

import { createHttpClient } from "@uma/toolkit/http"

const exampleAPI = createHttpClient({
  axios: { baseURL: "https://example.com" },
  rateLimit: { maxConcurrent: 1, minTime: 1200 },
  retry: { retries: 5 },
})

await exampleAPI.post("/api/v1/update", payload)

Every instance has its own Bottleneck limiter and axios‑retry strategy, so different services can run with different quotas.

Options reference

| Option | Default | Notes | | --------------------------- | ------------------- | ----------------------------------- | | axios.baseURL | "" | Standard Axios config | | axios.timeout | 10 000 | ms | | rateLimit.maxConcurrent | 5 | Parallel inflight cap | | rateLimit.minTime | 200 | ms gap between jobs | | retry.retries | 3 | Total attempts = 1 + retries | | retry.baseDelayMs | 100 | First back‑off step | | retry.maxJitterMs | 1000 | Randomised to avoid thundering herd | | retry.retryCondition(err) | network / 5xx / 429 | Override for fine‑grained control |