eschaq/plainlode

Plainlode is a market intelligence tool built for small owners of WooCommerce online stores. Instead of showing charts and dashboards, it gives you one plain-language decision about what to stock, based on live demand data. You type a store category, such as back to school, and Plainlode pulls real-time Google Trends interest data for that category through a service called Scrapingdog. It then runs that data through two AI model calls hosted on Fireworks AI, served on AMD Instinct GPUs. The first call is a cheap filter that keeps only the terms relevant to a stocking decision, and the second is a briefing engine that writes up findings, options, and a recommendation in plain language. A notable design choice is that Plainlode never makes up data. All numbers come from a live pull on each scan. If that live pull fails because of a rate limit or an upstream problem, it falls back to the most recent real snapshot it saved to disk and clearly labels it as a snapshot rather than live data. If there is neither live data nor a snapshot available, it says so plainly instead of guessing. Each recommendation also names one specific, watchable signal that would reverse the call, called the kill signal, so the store owner knows exactly what to monitor and when to change course. The project has a FastAPI backend that handles the trend pull, ranking, filtering, and briefing steps, and a React and Vite frontend with three simple states: input, scanning, and result. Both parts deploy as separate services on Railway. It requires a Scrapingdog API key and a Fireworks AI key to run, and is released under the MIT license, which allows free use including commercial use.