Show HN: I created an open source locally usable full fledged AI platform

An open source AI platform has been created that allows users to run AI models locally on their own infrastructure. The platform, called ENZO, offers a full-featured workspace that provides users with access to 300+ AI models. Users can build agents that write their own operating manuals and generate code, which they can then run on their own infrastructure.

The installation process is straightforward, with no accounts, mandatory environments, or databases required. Users simply open the app, press Login, and select a provider key. The keys are encrypted in the user's browser and saved in a passphrase-protected vault. If the user loses the passphrase, they can download a recovery file to regain access to their keys.

ENZO supports a wide range of providers, including OpenRouter, Google AI Studio, NVIDIA NIM, and Hugging Face, all of which offer free tiers. The platform ensures that provider keys never touch the server, as they are sealed in the user's browser vault. When a fresh self-hosted instance is created, the first live-validated key claimed by the instance is written to the container .env and sealed into the enzo-memory volume, ensuring immediate activation of server-side features such as agents, skills, and memory.

To get started, users can visit the hosted demo at https://enzo-hub.duckdns.org to try the platform in action. For those who prefer a self-hosted solution, the GitHub repository contains all the necessary code and configuration files. The platform requires Docker Desktop for Mac or Windows with WSL 2, allocating at least 4 GB of RAM. Users can then open the app, press Login, and paste a provider key to start using ENZO.