Profile My skills Saved Collections Edit profile Submit a skill

← all publishers

bentoml

@bentoml source repo

5 published skills

  1. ▌
    Bentoml Ec2 Deploy · bentoml bundle
    Deploy a containerized BentoML service directly onto one or more plain AWS EC2 instances with Docker — no Kubernetes. Takes a pushed container image (built by the bentoml-containerize skill), either uses the user's existing instances over SSH or provisions a new instance via the AWS CLI (SSM AMI lookup, security group, key pair), runs the container with restart-on-reboot, and verifies with a real inference request. Use when the user says things like "deploy my BentoML service to EC2", "run my bento on an AWS VM", "deploy this bento image to an EC2 instance", "run my BentoML container on AWS without Kubernetes", or "put my bento on a cloud VM". For Kubernetes targets use bentoml-k8s-deploy instead.
    0
    installs
  2. ▌
    Bentoml K8S Deploy · bentoml bundle
    Deploy a containerized BentoML service to a vanilla Kubernetes cluster using plain kubectl manifests (no Helm, no operators, no BentoCloud/Yatai). Takes a pushed container image (from the bentoml-containerize skill), discovers the bento's service topology, writes one `config.yml` for the deployment, renders one Deployment + Service per BentoML service (plus optional HPA/Ingress) from it, applies them in dependency order, and verifies the rollout with a real inference request. Use when the user says things like "deploy my BentoML service to Kubernetes", "deploy this bento image to my cluster", "run my bento on k8s", "create k8s manifests for my bento", "split my bento services into separate pods", or "expose my BentoML service in Kubernetes". Also diagnoses a deployment that went wrong — "my BentoML pods are crashing", "ImagePullBackOff", "pod stuck Pending", "readiness probe failing", "rollout stuck", "can't reach my service on Kubernetes", "inference 4xx/5xx".
    0
    installs
  3. ▌
    Bentoml Containerize · bentoml bundle
    Build a local BentoML project into a Bento, containerize it into an OCI/Docker image, smoke-test it locally, and push it to a container registry (Docker Hub, GHCR, ECR, private registry, kind/minikube local load, or ttl.sh). Use when the user asks to "containerize a Bento", "build a Docker image for my BentoML service", "package my BentoML service for deployment", "push my Bento image to a registry", or as the first step of deploying BentoML to Kubernetes or EC2. Does NOT deploy anything itself — hand off to bentoml-k8s-deploy or bentoml-ec2-deploy for that.
    0
    installs
  4. ▌
    Bentoml Create Bento · bentoml bundle
    Create a BentoML project: the `service.py` whose typed `@bentoml.api` methods become an HTTP API, plus its runtime image and a built Bento. Writes one from scratch, or converts existing code — a script, a notebook, a FastAPI/Flask app, an MLflow model, a BentoML 1.1 Runner project. Use for "create a bento", "wrap my model in BentoML", "convert my FastAPI app to BentoML", "add an endpoint to my bento", or a bento that will not start, rejects valid input, or serves one request at a time.
    0
    installs
  5. ▌
    Bentoml Deploy Scriptgen · bentoml bundle
    Generate a standalone, committable production deploy-script bundle (deploy/deploy.py + one config.yml — overrides only — from which the Kubernetes manifests are rendered) that builds, containerizes, pushes, deploys, and verifies a BentoML service without any agent involved — runnable from a terminal or CI/CD. Use when the user says things like "generate a deployment script", "deploy from CI/CD", "set up a production deployment pipeline", "deploy without the agent", "give me a script I can commit to deploy this", or "automate my BentoML deploys". Complements the interactive skills: bentoml-containerize, bentoml-k8s-deploy, and bentoml-ec2-deploy do a one-off deploy with you in the loop; this skill emits scripts that repeat it forever. Kubernetes and EC2 targets.
    0
    installs