Bentoml Model Packager

DevOps & Security

bentoml-model-packager

Bentoml Model Packager is an Agent Skill for devops & security published in the a5c-ai/babysitter repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

BentoML skill for model packaging, serving, and containerization.

View on GitHub ↗

QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

Open your project in Codex.
Copy and paste the prompt below.
Review files and risks before approving any action.

Prompt to paste

Inspect the Agent Skill at https://github.com/a5c-ai/babysitter/blob/HEAD/library/specializations/data-science-ml/skills/bentoml-model-packager/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/bentoml-model-packager, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

bentoml-model-packager

Overview

BentoML skill for model packaging, serving, and containerization with support for multiple ML frameworks.

Capabilities

Bento creation and versioning
Multi-framework model support (sklearn, PyTorch, TensorFlow, etc.)
API endpoint definition with validation
Docker containerization
Kubernetes deployment YAML generation
Adaptive batching configuration
Model signatures and runners
Service composition

Target Processes

Model Deployment Pipeline with Canary Release
Model Training Pipeline
ML System Integration Testing

Tools and Libraries

BentoML
Docker
Kubernetes

Input Schema

{
  "type": "object",
  "required": ["action"],
  "properties": {
    "action": {
      "type": "string",
      "enum": ["save", "build", "serve", "containerize", "push", "list"],
      "description": "BentoML action to perform"
    },
    "modelConfig": {
      "type": "object",
      "properties": {
        "name": { "type": "string" },
        "framework": { "type": "string" },
        "modelPath": { "type": "string" },
        "signatures": { "type": "object" }
      }
    },
    "serviceConfig": {
      "type": "object",
      "properties": {
        "servicePath": { "type": "string" },
        "port": { "type": "integer" },
        "workers": { "type": "integer" },
        "batchConfig": {
          "type": "object",
          "properties": {
            "maxBatchSize": { "type": "integer" },
            "maxLatencyMs": { "type": "integer" }
          }
        }
      }
    },
    "buildConfig": {
      "type": "object",
      "properties": {
        "bentoName": { "type": "string" },
        "version": { "type": "string" },
        "includeFiles": { "type": "array", "items": { "type": "string" } },
        "pythonRequirements": { "type": "string" }
      }
    },
    "containerConfig": {
      "type": "object",
      "properties": {
        "imageName": { "type": "string" },
        "registry": { "type": "string" },
        "dockerOptions": { "type": "object" }
      }
    }
  }
}

Output Schema

{
  "type": "object",
  "required": ["status", "action"],
  "properties": {
    "status": {
      "type": "string",
      "enum": ["success", "error"]
    },
    "action": {
      "type": "string"
    },
    "modelTag": {
      "type": "string"
    },
    "bentoTag": {
      "type": "string"
    },
    "imageTag": {
      "type": "string"
    },
    "endpoint": {
      "type": "string"
    },
    "kubernetesYaml": {
      "type": "string"
    }
  }
}

Usage Example

{
  kind: 'skill',
  title: 'Package and containerize model',
  skill: {
    name: 'bentoml-model-packager',
    context: {
      action: 'containerize',
      modelConfig: {
        name: 'fraud_classifier',
        framework: 'sklearn',
        modelPath: 'models/fraud_model.pkl'
      },
      buildConfig: {
        bentoName: 'fraud-service',
        version: '1.0.0',
        pythonRequirements: 'requirements.txt'
      },
      containerConfig: {
        imageName: 'fraud-service',
        registry: 'gcr.io/my-project'
      }
    }
  }
}