بناء Bento وتحويله إلى حاوية

A Bento is the deployable bundle BentoML builds. It packs your service code, the model, and the exact dependencies into one archive. 📦

You describe the build in a bentofile.yaml. It names the service entry point and lists what to include.

You tell BentoML which source files to bundle. The include key takes globs so only the code you need ships.

Dependencies live right in the bentofile. Under python you list packages so the build resolves them into the bundle.

One command assembles everything. bentoml build reads the bentofile and produces a versioned Bento in your store.

Built Bentos show up alongside your models. Run bentoml list to see each one with its name and version tag.

A Bento still needs a runtime. Turning it into a Docker image makes it run identically on any machine or cloud. 🐳

BentoML writes the Dockerfile for you. bentoml containerize takes a Bento tag and builds a ready-to-run image.

Now it is a normal image. You start it with docker run, mapping the port so requests reach the service inside.

That image is your deployable artifact. Push it to a registry and any cloud runtime or Kubernetes cluster can pull and serve it. 🚀

You can control the runtime under your image. The docker section of the bentofile lets you pick a Python version or base.

You ran bentoml build and got a Bento. What does bentoml containerize add on top of that?

You defined a bentofile, ran bentoml build to bundle code and deps, then containerized it into a Docker image ready to ship to any cloud. End to end! 🙌