gopiy3k/growth-os

Growth OS, also called Engineering OS or EOS, is an open source platform for running AI powered jobs in a reliable, organized way. Instead of building AI logic straight into a product, other software can hand off a unit of work, called a job, to Growth OS. Growth OS puts that job into a durable queue, has a worker pick it up, reason about it using an AI call, and then save a structured result that other systems can read later. Nothing gets lost even if something fails partway through, since failed jobs move into a dead letter queue rather than disappearing. The project separates three concerns so each one can change independently. Growth OS itself, built on Supabase, owns the durable state: the queue, the failed job queue, and the stored results. A layer called Engineering OS handles orchestration, meaning scheduling, claiming jobs, retrying them, and making the process observable. A separate piece called the AI Runtime handles the actual reasoning step, meaning picking a model or provider and managing authentication, and it does this through a standard OpenAI-compatible chat completions interface rather than being tied to any specific AI company or model. This means the reasoning engine can be swapped out without changing how jobs are queued or tracked. The system is described as engine-agnostic, so while the first working example included is a Content engine that scores external items editorially, the same underlying queue and worker pattern is meant to support future engines for other tasks as well. Other products are expected to interact with Growth OS only through its public API and its events, and never by reading or writing its internal database tables directly, which keeps it decoupled from any one product's codebase. Getting started does not require any paid services or private infrastructure. The README walks through cloning the repository, setting up a Python virtual environment, and running a built in mock AI runtime locally so you can send a real job through the whole pipeline without needing an actual Supabase project or any AI provider credentials. The project is written in Python and released under the Apache License 2.0, which allows free use, modification, and redistribution, including for commercial purposes, as long as license and copyright notices are preserved.