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

# LLM Providers

> Configure which AI model powers your agent

Choose the AI model that powers your agent - OpenAI, Anthropic, local models, and more.

```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
graph LR
    subgraph "LLM Providers"
        A[🤖 Agent] --> L[💭 LLM Provider]
        L --> O[OpenAI]
        L --> AN[Anthropic]
        L --> OL[Ollama]
    end
    
    classDef agent fill:#6366F1,stroke:#7C90A0,color:#fff
    classDef llm fill:#F59E0B,stroke:#7C90A0,color:#fff
    classDef provider fill:#10B981,stroke:#7C90A0,color:#fff
    
    class A agent
    class L llm
    class O,AN,OL provider
```

## Quick Start

<Steps>
  <Step title="Use Default Model">
    ```rust theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
    use praisonai::Agent;

    // Uses gpt-4o-mini by default
    let agent = Agent::new()
        .name("Assistant")
        .instructions("You are helpful")
        .build()?;
    ```
  </Step>

  <Step title="Choose a Model">
    ```rust theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
    use praisonai::Agent;

    let agent = Agent::new()
        .name("Assistant")
        .model("gpt-4o")  // or "claude-3-opus", "ollama/llama3"
        .build()?;
    ```
  </Step>

  <Step title="Custom API Endpoint">
    ```rust theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
    use praisonai::Agent;

    let agent = Agent::new()
        .name("Assistant")
        .model("llama3")
        .base_url("http://localhost:11434/v1")
        .build()?;
    ```
  </Step>
</Steps>

***

## How It Works

```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
sequenceDiagram
    participant User
    participant Agent
    participant LLM
    
    User->>Agent: chat("Hello")
    Agent->>LLM: Messages + Tools
    LLM-->>Agent: Response
    Agent-->>User: "Hello! How can I help?"
```

***

## Choosing a Model

```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
graph TB
    subgraph "Which Model?"
        Q{Need?} --> |Speed| F[gpt-4o-mini]
        Q --> |Quality| G[gpt-4o]
        Q --> |Privacy| O[Ollama/Local]
        Q --> |Long context| C[claude-3]
    end
    
    classDef question fill:#6366F1,stroke:#7C90A0,color:#fff
    classDef model fill:#10B981,stroke:#7C90A0,color:#fff
    
    class Q question
    class F,G,O,C model
```

| Model           | Best For                       | Speed     |
| --------------- | ------------------------------ | --------- |
| `gpt-4o-mini`   | Fast responses, cost effective | ⚡ Fast    |
| `gpt-4o`        | High quality, complex tasks    | 🔵 Medium |
| `claude-3-opus` | Long documents, analysis       | 🔵 Medium |
| `ollama/llama3` | Privacy, offline use           | 🟢 Local  |

***

## Configuration

| Option        | Type     | Default        | Description         |
| ------------- | -------- | -------------- | ------------------- |
| `model`       | `String` | `gpt-4o-mini`  | Model name          |
| `api_key`     | `String` | From ENV       | API key             |
| `base_url`    | `String` | OpenAI default | API endpoint        |
| `temperature` | `f32`    | `0.7`          | Randomness (0-1)    |
| `max_tokens`  | `u32`    | None           | Max response length |

***

## Best Practices

<AccordionGroup>
  <Accordion title="Use environment variables for API keys">
    Set `OPENAI_API_KEY` instead of hardcoding keys.
  </Accordion>

  <Accordion title="Start with gpt-4o-mini">
    Fast and cheap - upgrade to gpt-4o only when needed.
  </Accordion>

  <Accordion title="Use local models for privacy">
    Ollama runs models locally with no data leaving your machine.
  </Accordion>
</AccordionGroup>

***

## Related

<CardGroup cols={2}>
  <Card title="Agent" icon="robot" href="/docs/rust/agent">
    Agent configuration
  </Card>

  <Card title="Streaming" icon="stream" href="/docs/rust/streaming">
    Stream responses
  </Card>
</CardGroup>