Kamil Jopek kamilio

Inkling reasoning effort through Vercel AI Gateway

Neither reasoning_effort nor Vercel's reasoning.effort object shows reliable effort scaling for thinkingmachines/inkling through Vercel AI Gateway.

Thinking Machines documents reasoning_effort for its OpenAI-compatible Chat Completions API and maps low to 0.2 and xhigh to 0.99. It also warns that individual samples are stochastic, so this reproduction compares multiple samples with enough output headroom to avoid truncation.

How to run

npm init -y && npm install ai @ai-sdk/openai zod tsx
POE_API_KEY=your-key npx tsx poe-repro.ts

Bug

/v1/responses returns invalid output_text for Google models when making tool calls:

How to configure gems

Are you writing a new gem and you are not sure how to let user configure it? Here is a short gist to show you how to do it. I like this style of configurations, inspired by many already existing gems such as Devise, Airbrake, PDFkit and many more.

# Configuration
MyShinyGem.configure do |config|
  config.binary = '/bin/ls'
 config.username = 'superuser'

	// npm install openai
	import OpenAI from "openai";

	const client = new OpenAI({
	apiKey: process.env.VERCEL_API_KEY,
	baseURL: "https://ai-gateway.vercel.sh/v1",
	});

	const completion = await client.chat.completions.create({
	model: "moonshotai/kimi-k3",

	const response = await fetch("https://ai-gateway.vercel.sh/v1/responses", {
	method: "POST",
	headers: {
	Authorization: `Bearer ${process.env.VERCEL_API_KEY}`,
	"Content-Type": "application/json",
	},
	body: JSON.stringify({
	model: "meta/muse-spark-1.1",
	input: "What are today's headlines?",
	reasoning: { effort: "high", summary: "detailed" },

	{
	"name": "investigation-scripts",
	"version": "1.0.0",
	"description": "",
	"main": "test_poe_opus46_anthropic.js",
	"scripts": {
	"test": "echo \"Error: no test specified\" && exit 1"
	},
	"keywords": [],
	"author": "",

	// Repro: AI SDK (OpenAI Responses) calling Poe API (google/nano-banana)
	//
	// Install (in an empty repo):
	// npm i ai @ai-sdk/openai
	//
	// Run:
	// node --env-file=.env tests/scripts/repro-responses-google-nano-banana.mjs
	// or:
	// POE_API_KEY=your_key node tests/scripts/repro-responses-google-nano-banana.mjs

	// Test prompt caching with Anthropic SDK via Poe API
	//
	// Run:
	// node --env-file=.env tests/scripts/test-prompt-caching.mjs
	// or:
	// POE_API_KEY=your_key node tests/scripts/test-prompt-caching.mjs

	import Anthropic from "@anthropic-ai/sdk";

	const client = new Anthropic({

	// *** Middleware - custom thunk implementation (slightly diffferent though) - maybe more scalable appraoch
	export default function createEventBusMiddleware(eventBus) {
	return ({ dispatch, getState }) => (next) => (action) => {
	// we could optimize further here and use rest operator to get all arguments and compose the middlewares
	if (typeof action === 'function') {
	return action({ dispatch, eventBus, getState });
	}

	return next(action);
	};

	// ***************************
	// Programming with functions (Kamil Jopek)
	// Stuttgart JS Meetup Talk (13th April 2015)
	// ***************************

	// Basic functions
	var add = function(a, b) {
	return a + b;
	};