Show HN: LLM Newsletter Kit – Automate expert newsletters for $0.20/issue (github.com)

LLM Newsletter Kit is a TypeScript, type‑first framework that automates end‑to‑end newsletter production with an explicit Crawling → Analysis → Content Generation → Save pipeline. Built as a dependency‑injection toolkit, every stage (crawlers, LLMs, DB, logging, email) is swappable via provider interfaces, letting teams plug in custom scrapers, models, or databases without changing the orchestration. It targets production use: Node ≥22, Rollup outputs (ESM+CJS+d.ts), Vitest with 100% coverage, GitHub Actions CI, retries/chain options, preview email support, and orchestration via @langchain/core runnables. Why it matters: the kit makes advanced AI workflows practical for newsletter publishers and researchers—supporting multi‑step reasoning (self‑reflection, chain‑of‑thought), per‑stage model selection, token caps, and granular retry/cost controls that no‑code platforms struggle to provide. The author extracted the core from Research Radar (a real Korean cultural‑heritage service) where it ran 24/7 with near‑zero maintenance, 15% CTR, and per‑issue LLM costs of ~$0.20–$1. The trade‑off is higher setup complexity than no‑code tools, but strong type contracts, IDE autocompletion, and a full reference implementation (including Prisma/Drizzle integrations, HTML email templates, and example OpenAI model wiring) make it a pragmatic, production‑grade starting point for building customizable, cost‑controlled AI media pipelines.