Introduction
Deployable AI is a small toolkit for serving local model inference quickly. It follows a familiar model-package pattern: register a model artifact, provide an inference script, and expose the model through a predictable API endpoint.
Workflow
The repo centers on three pieces:
- A serialized model artifact, usually saved as a
.joblibfile - An inference script with
input_fnandpredict_fn - A serving command that starts a local backend with one endpoint per registered model
Notes
The project is intentionally narrow: it targets JSON request workflows and local development speed. It is useful as a compact serving layer when the goal is to validate model behavior before investing in heavier deployment infrastructure.