> ## Documentation Index
> Fetch the complete documentation index at: https://cyborg-encryption-copy.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Deployment Models

CyborgDB offers two deployment approaches, each optimized for different use cases and operational requirements. Both models are **self-hosted** and work with your existing database infrastructure.

***

## CyborgDB Service

**Self-deployed REST API service**

Deploy CyborgDB as a standalone microservice that provides REST API access to confidential vector search. The service runs on your infrastructure and can scale independently from your applications.

<Tip>Learn more about CyborgDB Service [here](../service/guides/intro/about).</Tip>

### Key Benefits

**Independent Scaling**\
Scale vector operations separately from your main application. Handle high query loads without impacting your core services.

**Self-Optimization**\
The service automatically optimizes index performance, manages memory efficiently, and adapts to query patterns over time.

**Multi-Language Access**\
One service deployment supports multiple applications and programming languages through REST API and client SDKs.

**Operational Simplicity**\
Centralized deployment, monitoring, and maintenance. Update vector search capabilities without touching application code.

### Deployment Options

<CardGroup cols={2}>
  <Card title="Docker Deployment" href="../service/guides/intro/quickstart-docker" icon="docker">
    *Containerized service*

    Deploy as a Docker container for easy orchestration and scaling
  </Card>

  <Card title="Python Service" href="../service/guides/intro/quickstart-python" icon="python">
    *Direct Python installation*

    Install and run as a Python service with pip for lightweight deployment
  </Card>
</CardGroup>

### Client SDKs

<CardGroup cols={3}>
  <Card title="Python SDK" href="../service/python-sdk/introduction" icon="python">
    Full-featured Python client with async support and type hints
  </Card>

  <Card title="JavaScript/TypeScript SDK" href="../service/js-ts-sdk/introduction" icon="js">
    Modern JS/TS client with Promise-based API and TypeScript definitions
  </Card>

  <Card title="Go SDK" href="../service/go-sdk/introduction" icon="golang">
    High-performance Go client with idiomatic API and concurrency support
  </Card>
</CardGroup>

***

## CyborgDB Embedded

**Direct library integration**

Embed CyborgDB directly into your applications using Python or C++ libraries. This approach provides maximum control and performance by eliminating network overhead.

<Tip>Learn more about CyborgDB Embedded [here](../embedded/guides/intro/about).</Tip>

### Key Benefits

**Maximum Performance**\
Direct memory access and zero network latency. Ideal for latency-sensitive applications requiring sub-millisecond response times.

**Deep Integration**\
Customize every aspect of vector operations. Perfect for specialized workflows and performance tuning requirements.

**Complete Control**\
Full ownership of the vector search stack. No external dependencies or service management overhead.

**Advanced Customization**\
Access to low-level APIs for custom index configurations, memory management, and algorithm tuning.

### Deployment Options

<CardGroup cols={2}>
  <Card title="Python Embedded" href="../embedded/python/introduction" icon="python">
    *Direct Python integration*

    Embed confidential vector search directly in Python applications
  </Card>

  <Card title="C++ Embedded" href="../embedded/cpp/introduction" icon="brackets-curly">
    *Native C++ integration*

    High-performance native integration for C++ applications
  </Card>
</CardGroup>

***

## Choosing the Right Model

<Tabs>
  <Tab title="Use CyborgDB Service">
    * **Multi-application architecture** - Multiple services need vector search capabilities
    * **Team scalability** - Different teams use different programming languages
    * **Operational simplicity** - You want centralized vector search management
    * **Independent scaling** - Vector workloads need to scale separately from applications
    * **Microservice patterns** - You prefer service-oriented architecture
  </Tab>

  <Tab title="Embedded Deployment">
    * **Maximum performance** - Sub-millisecond latency requirements
    * **Deep customization** - Need fine-grained control over vector operations
    * **Single application focus** - One primary application using vector search
    * **Resource optimization** - Want to eliminate network overhead entirely
    * **Specialized workflows** - Require custom integration or algorithm tuning
  </Tab>
</Tabs>

Start with embedded libraries for rapid prototyping, then scale to service deployment for production multi-application environments, or vice versa based on your specific requirements.