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

# LLM Providers

> Choose your AI's brain. From lightning-fast Groq to privacy-focused local models, Vizra ADK supports 10 different LLM providers via Prism PHP. Find the perfect match for your use case.

<Info>
  Not all AI models are created equal. Some excel at creative writing, others at coding, and some are just blazing fast. With Vizra ADK, you can mix and match providers to give each agent the perfect brain for its job.
</Info>

## Quick Setup

Getting started is as easy as 1-2-3.

### 1. Set Your Default Provider

```bash .env theme={null}
VIZRA_ADK_DEFAULT_PROVIDER=openai
VIZRA_ADK_DEFAULT_MODEL=gpt-4-turbo

# Don't forget your API key!
OPENAI_API_KEY=sk-...
```

### 2. Or Configure Per Agent

```php app/Agents/CustomerSupportAgent.php theme={null}
use Prism\Prism\Enums\Provider;

class CustomerSupportAgent extends BaseLlmAgent
{
    protected ?Provider $provider = Provider::Anthropic;
    protected string $model = 'claude-3-opus-20240229';
}
```

### 3. Or Switch on the Fly

```php theme={null}
$agent->setProvider('groq')
    ->setModel('mixtral-8x7b-32768')
    ->setTemperature(0.7);
```

## Supported Providers

Vizra ADK supports **10 LLM providers** via Prism PHP. Each brings something special to the table.

<AccordionGroup>
  <Accordion title="OpenAI" icon="circle-check" defaultOpen>
    The industry leader - GPT-4 is your Swiss Army knife for any task.

    | Best For                                         | Popular Models                         |
    | ------------------------------------------------ | -------------------------------------- |
    | General purpose, creative tasks, code generation | gpt-4-turbo, gpt-3.5-turbo, o1-preview |

    **Environment Variable:** `OPENAI_API_KEY`

    ```php theme={null}
    protected ?Provider $provider = Provider::OpenAI;
    protected string $model = 'gpt-4-turbo';
    ```
  </Accordion>

  <Accordion title="Anthropic (Claude)" icon="circle-check">
    Claude is thoughtful, careful, and great at complex reasoning.

    | Best For                                    | Popular Models                 |
    | ------------------------------------------- | ------------------------------ |
    | Analysis, writing, safe & ethical responses | claude-3-opus, claude-3-sonnet |

    **Environment Variable:** `ANTHROPIC_API_KEY`

    ```php theme={null}
    protected ?Provider $provider = Provider::Anthropic;
    protected string $model = 'claude-3-opus-20240229';
    ```
  </Accordion>

  <Accordion title="Google Gemini" icon="circle-check">
    Multimodal magic - handles text, images, and more.

    | Best For                                          | Popular Models                   |
    | ------------------------------------------------- | -------------------------------- |
    | Multimodal tasks, fast responses, cost efficiency | gemini-1.5-pro, gemini-1.5-flash |

    **Environment Variable:** `GEMINI_API_KEY`

    ```php theme={null}
    protected ?Provider $provider = Provider::Gemini;
    protected string $model = 'gemini-1.5-flash';
    ```
  </Accordion>

  <Accordion title="Groq" icon="bolt">
    Speed demon. Insanely fast inference for real-time applications.

    | Best For                          | Popular Models                      |
    | --------------------------------- | ----------------------------------- |
    | Real-time chat, low latency needs | mixtral-8x7b-32768, llama2-70b-4096 |

    **Environment Variable:** `GROQ_API_KEY`

    ```php theme={null}
    protected ?Provider $provider = Provider::Groq;
    protected string $model = 'mixtral-8x7b-32768';
    ```
  </Accordion>

  <Accordion title="Ollama (Local)" icon="server">
    Run models locally - your data never leaves your machine.

    | Best For                              | Popular Models                  |
    | ------------------------------------- | ------------------------------- |
    | Privacy, offline use, experimentation | llama2, codellama, mistral, phi |

    ```php theme={null}
    protected ?Provider $provider = Provider::Ollama;
    protected string $model = 'llama2';
    ```

    <Warning>
      Requires Ollama to be installed and running locally.
    </Warning>
  </Accordion>

  <Accordion title="OpenRouter" icon="shuffle">
    One API, 100+ models - the ultimate flexibility.

    | Best For                                          | Available Models                            |
    | ------------------------------------------------- | ------------------------------------------- |
    | Model flexibility, A/B testing, cost optimization | GPT-4, Claude, Gemini, Llama, and 100+ more |

    **Environment Variable:** `OPENROUTER_API_KEY`

    ```php theme={null}
    protected ?Provider $provider = Provider::OpenRouter;
    protected string $model = 'openai/gpt-4-turbo';
    // Or use any model: 'anthropic/claude-3-opus', 'google/gemini-pro', etc.
    ```

    <Tip>
      Perfect for testing different models without managing multiple API keys.
    </Tip>
  </Accordion>

  <Accordion title="DeepSeek" icon="code">
    Specialized models for chat and code generation.

    **Environment Variable:** `DEEPSEEK_API_KEY`

    ```php theme={null}
    protected ?Provider $provider = Provider::DeepSeek;
    protected string $model = 'deepseek-coder';
    ```
  </Accordion>

  <Accordion title="Mistral AI" icon="wind">
    Open-weight models with impressive performance.

    **Environment Variable:** `MISTRAL_API_KEY`

    ```php theme={null}
    protected ?Provider $provider = Provider::Mistral;
    protected string $model = 'mistral-large-latest';
    ```
  </Accordion>

  <Accordion title="xAI (Grok)" icon="x">
    Grok models from Elon's xAI team.

    **Environment Variable:** `XAI_API_KEY`

    ```php theme={null}
    protected ?Provider $provider = Provider::XAI;
    protected string $model = 'grok-beta';
    ```
  </Accordion>

  <Accordion title="Voyage AI" icon="compass">
    Embedding specialist for semantic search.

    **Environment Variable:** `VOYAGEAI_API_KEY`

    ```php theme={null}
    protected ?Provider $provider = Provider::VoyageAI;
    protected string $model = 'voyage-large-2';
    ```

    <Warning>
      Primarily for embeddings, not text generation.
    </Warning>
  </Accordion>
</AccordionGroup>

## Cool Provider Tricks

<CardGroup cols={2}>
  <Card title="Auto-Detection Magic" icon="wand-magic-sparkles">
    Just set the model name and Vizra figures out the provider.

    ```php theme={null}
    // These auto-detect the provider!
    protected string $model = 'gpt-4';        // OpenAI
    protected string $model = 'claude-3-opus'; // Anthropic
    protected string $model = 'gemini-pro';    // Gemini
    protected string $model = 'llama2';        // Ollama
    ```
  </Card>

  <Card title="Runtime Switching" icon="arrows-rotate">
    Change providers on the fly based on user needs.

    ```php theme={null}
    // Premium users get the good stuff!
    if ($user->isPremium()) {
        $agent->setProvider('anthropic')
              ->setModel('claude-3-opus-20240229');
    } else {
        $agent->setProvider('openai')
              ->setModel('gpt-3.5-turbo');
    }
    ```
  </Card>
</CardGroup>

## Choosing the Right Provider

Not sure which to pick? Here's your cheat sheet.

| Use Case         | Recommended Provider                                      |
| ---------------- | --------------------------------------------------------- |
| Business Apps    | OpenAI GPT-4 or Anthropic Claude - reliable and versatile |
| Speed Demons     | Groq or Gemini Flash - when milliseconds matter           |
| Privacy First    | Ollama - keep everything local and secure                 |
| Code Generation  | DeepSeek Coder or GPT-4 - they speak fluent code          |
| Long Context     | Claude 3 or Gemini 1.5 Pro - handle entire books          |
| Budget Conscious | GPT-3.5 Turbo or local Ollama models                      |

## API Keys Setup

Don't forget to add your API keys to the `.env` file.

```bash .env theme={null}
# OpenAI
OPENAI_API_KEY=sk-...

# Anthropic
ANTHROPIC_API_KEY=sk-ant-...

# Google Gemini
GEMINI_API_KEY=...

# OpenRouter (100+ models!)
OPENROUTER_API_KEY=...

# Add others as needed!
GROQ_API_KEY=...
MISTRAL_API_KEY=...
# etc...
```

## Next Steps

Now that you've picked your provider, let's create some amazing agents.

<Card title="Create Your First Agent" icon="robot" href="/concepts/agents">
  Learn how to build powerful AI agents with your chosen provider.
</Card>