Welcome to Baseten!
Bring your models. We'll handle the rest.
Baseten is a hosted platform for building ML-powered applications. Build apps with auto-scaling, GPU access, CRON jobs, and serverless functions.
Read our getting started guide for an overview of the platform and an opinionated reference architecture.
Welcome to Baseten
We work hard to create and maintain this comprehensive, readable, and occasionally even personable reference to assist you in building full-stack applications powered by ML models. If anything is unclear or mysteriously broken, we want to help! You can get in touch with us via the support button on the lower-right corner of every docs page or by sending an email to [email protected]. We truly appreciate feedback on the docs themselves as well.
These docs are organized into five sections: models, applications, data, settings, and an appendix.
Start by getting your ML models out of a Jupyter notebook and on to robust production infrastructure. Serve, manage, and monitor your own models with just a few lines of Python, or kickstart a project with pre-trained models. Model deployment workflows are powered by our open-source library, Truss.
Integrate your machine learning models into full-stack applications with custom business logic and powerful user interfaces—no frontend, backend, or MLOps knowledge required. You can start from scratch or heat up your project with a starter application powered by a pre-trained model. Either way, you'll be dishing out piping-hot ML-powered applications in no time.
Let's stack some bytes. This is where the data for all of your applications and models lives. Create tables in your Baseten Postgres database; connect to external data sources like MySQL, BigQuery, Snowflake, and Redis; and write queries to populate your applications.
Anything you need for managing your account is explained here. Baseten is even more fun with friends; learn how to invite and manage other users within your workspace. Explore tips, tricks, and tactics for making your Baseten experience a tailored fit.
We understand that this documentation isn't a novel; no one is sitting down to read it cover-to-cover. Still, we strive for a level of narrative flow within the documentation. API references and lists of terms interrupt poetic prose like boulders in a mountain stream. Thus, the appendix.