Deployment
Runs where your work runs.
foundation4 is small, self-contained and works with the models you already have.
Built for isolated networks
foundation4 is designed to run on networks with no outside connection. Some models and the token splitter download on first use, so the installation guideshows how to load them ahead of time for a disconnected install.
Light enough to accredit
foundation4 needs one PostgreSQL database and runs on CPUs. It installs with two Helm charts.
Fewer moving parts means less to review, less to patch and less to explain to the people who approve what runs on your network.
Use the models you already run
foundation4 works with any OpenAI-compatible model. Point it at the model servers you already operate and keep the choice of model yours.
Platform
Security
Your content stays inside your perimeter, encrypted with a separate key for each classification.
Platform
Deployment
Built for isolated networks, light enough to accredit, and works with the models you already run.
Platform
Search and API
Search by meaning and by exact words, through a REST API ready for production integration.
Talk to us
Get a briefing for your team, or read the docs to go deeper.