Skip to main content
This recipe is the SDK-only baseline: collect context, rank candidates deterministically, and hand a compact packet to a human or model layer. It does not trade.

​
Install

npm init -y
npm install @spfunctions/sdk@1.0.1
export SF_API_KEY="sf_..."

​
Script

import { writeFile } from "node:fs/promises"
import { SimpleFunctions } from "@spfunctions/sdk"

const sf = new SimpleFunctions({
  baseUrl: "https://simplefunctions.dev",
  apiKey: process.env.SF_API_KEY,
})

const world = await sf.world.get()
const screen = await sf.intelligence.screen({
  venue: "kalshi",
  volMin: 100,
  sort: "volume",
  order: "desc",
  limit: 12,
  nextActions: false,
})

const inspected = []
for (const market of screen.markets.slice(0, 5)) {
  inspected.push(await sf.markets.get(market.ticker, { nextActions: false }))
}

const ranked = inspected
  .map(market => ({
    ticker: market.ticker,
    title: market.title,
    price: market.price,
    bestBid: market.bestBid,
    bestAsk: market.bestAsk,
    spread: market.spread,
    volume24h: market.volume24h,
    closeTime: market.closeTime,
    score:
      Math.log10(1 + Number(market.volume24h ?? 0)) * 20 -
      Number(market.spread ?? 10) * 4,
  }))
  .sort((a, b) => b.score - a.score)

const packet = {
  generatedAt: new Date().toISOString(),
  worldAsOf: world.asOf,
  regime: world.regime?.label,
  candidates: ranked.slice(0, 5),
}

await writeFile("research-packet.json", JSON.stringify(packet, null, 2))
console.log(JSON.stringify(packet, null, 2))

​
Production notes

  • Keep the scoring deterministic before adding a model.
  • Use nextActions: false when you want compact machine packets.
  • Inspect before acting on any ticker from a screener.
  • Store world.asOf and the source request ids with downstream decisions.

​
Next steps

SDK quickstart

Install and first calls.

Agent market monitor

Add a model loop after the deterministic screen.