Francisco Moretti FranciscoMoretti

NPM releases for a pnpm monorepo (GitHub Actions + Changesets)

Minimal, reliable setup to auto-version, create GitHub Releases, and publish to npm from main using Changesets. No GitHub App required (uses GITHUB_TOKEN).

What this does

Opens a "Version Packages" PR when changesets land on main.
On merging that PR, publishes to npm and creates GitHub Releases.
Optionally deploys to Vercel after publish (see VERCEL_DEPLOYMENTS.md).

NPM releases for a pnpm monorepo (GitHub Actions + Changesets)

Minimal, reliable setup to auto-version, create GitHub Releases, and publish to npm from main using Changesets. No GitHub App required (uses GITHUB_TOKEN).

What this does

Opens a "Version Packages" PR when changesets land on main.
On merging that PR, publishes to npm and creates GitHub Releases.
Optionally deploys to Vercel after publish (see VERCEL_DEPLOYMENTS.md).

Goal: Turn my current working tree diff into a PR to upstream. Infer upstream_repo, branch_name, commit_message, PR title/body from git remotes, git status, and my message.

My message (use this to infer type/scope/subject and PR body): """ <WRITE YOUR MESSAGE HERE — first line becomes the Conventional Commit subject, lines after a blank line become the body> """

Behavior:

Non-interactive. Assume gh is installed and authenticated. Use | cat to avoid pagers.

Create a Deep Research Tool (AI SDK)

In this guide, you’ll build a Deep Research tool using AI SDK. It orchestrates two agents — a Supervisor and a Researcher — and runs through three concrete phases:

Clarifying questions (optional but automatic)
Research orchestration (supervisor) and evidence gathering/compression (researcher)
Final report writing with citations

What you’ll get

Feature	fs.createWriteStream(path).write(buffer)	fs.writeFileSync(path, buffer)
Synchronicity	Asynchronous	Synchronous
Blocking	Non-blocking	Blocking
Performance	Better for large files or frequent writes	Better for small, infrequent writes
Memory Usage	More memory-efficient for large files	Loads entire content into memory
Error Handling	Requires callback or event listeners	Throws errors directly
Use Case	Streaming large files, real-time data	Quick, simple writes, small files
Control Flow	Continues execution immediately	Waits until write is complete
Backpressure Handling	Handles backpressure autom

	I want you to fix mypy errors

	To do so, you need to do it by pieces.

	The best way to do it is to solve by file.

	Here are 2 important commands:

	1) Summary of number of mypy errors per file

	I want you to fix biome violations

	To do so, you need to do it by pieces
	The best way to do it is to solve by rule

	Here are 2 important commands:
	```
	npx biome check . --reporter=summary
	```
	Will print a summary of number of violations per file


	const models = [
	{
	id: 'openai/gpt-5',
	provider: 'openai',
	},
	{
	id: 'openai/gpt-4o',
	provider: 'openai',
	},

	#!/usr/bin/env tsx
	import { mkdirSync, writeFileSync } from "node:fs";
	import { dirname, join } from "node:path";
	import { fileURLToPath } from "node:url";
	import {
	type AiGatewayModelsResponse,
	AiGatewayModelsResponseSchema,
	} from "../../packages/vercel-gateway/ai-sdk-models-schemas";

	function ensureDir(filePath: string) {

	import { create } from 'zustand';
	import { subscribeWithSelector, devtools } from 'zustand/middleware';
	import {
	AbstractChat,
	type ChatInit,
	type ChatState,
	type ChatStatus,
	type UIMessage,
	} from 'ai';
	import type { ChatMessage } from '@/lib/ai/types';