> ## Documentation Index
> Fetch the complete documentation index at: https://docs.budecosystem.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Deployment Concepts

> Understand endpoint usage, publishing, tabs, and lifecycle controls

## What is a Deployment?

A deployment is a managed runtime endpoint that binds a selected model to infrastructure, execution settings, and access controls. It is the unit used for serving inference traffic in Bud AI Foundry projects.

## Deploy vs Use vs Publish

* **Deploy**: Creates an endpoint and makes it available for inference once active.
* **Use this model**: Gives ready-to-copy cURL, Python, and JavaScript snippets to call that endpoint.
* **Publish**: Lists the model in the **Customer Dashboard portal** for customer-facing consumption and pricing governance.

## Deployment Building Blocks

```mermaid theme={null}
flowchart LR
    A[Model Source] --> D[Deployment]
    B[Cluster Target] --> D
    C[Runtime Config] --> D
    D --> E[Active Endpoint]
    E --> F[Use this model Snippets]
    E --> G[Optional Publish to Customer Portal]
```

## Deployment Types

### Cloud Deployments

Use managed cloud model providers when you need fast setup and external model access.

### Local Deployments

Use local model artifacts (for example Hugging Face or disk-based assets) when you need infrastructure control or custom runtime tuning.

## Deployment Detail Tabs

### General

Shows model, cluster, and status-level summary information.

### Workers

Available for local deployments. Shows worker state, placement, and capacity signals.

### Settings

Central place to configure rate limits, retries, and fallback behavior.

## Lifecycle States

```mermaid theme={null}
stateDiagram-v2
    [*] --> Creating
    Creating --> Active
    Active --> UsableByAPIClients
    UsableByAPIClients --> PublishedToCustomerPortal
    PublishedToCustomerPortal --> Unpublished
    Unpublished --> PublishedToCustomerPortal
    Active --> Failed
    Failed --> Active
    Active --> Deleted
    Unpublished --> Deleted
    Deleted --> [*]
```

## Reliability Concepts

* **Rate Limiting** controls traffic volume and burst behavior.
* **Retry Limits** define automatic re-attempt behavior.
* **Fallback Chains** route traffic to alternate endpoints during failure conditions.

## Permission Model

Deployment actions are permission-aware. View-only users can inspect metadata, while manage permissions are required for create, edit, publish, and delete actions.