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

# LLM Providers

> Configure OpenAI, Anthropic, Gemini, or Ollama for enhanced analysis

# LLM Providers

Dorgu can use large language models to enhance its application analysis. LLM integration provides deeper code understanding, framework-specific recommendations, smarter resource sizing, and security best practices tailored to your stack.

<Info>
  LLM enhancement is **optional**. Dorgu's static analysis and manifest generation work without any LLM provider configured. The LLM layer adds refinements on top of the base analysis.
</Info>

## Supported Providers

| Provider  | Models                   | Env Variable                         | Default Model   |
| --------- | ------------------------ | ------------------------------------ | --------------- |
| OpenAI    | gpt-4, gpt-3.5-turbo     | `OPENAI_API_KEY`                     | gpt-4           |
| Anthropic | claude-3-sonnet          | `ANTHROPIC_API_KEY`                  | claude-3-sonnet |
| Gemini    | gemini-1.5-pro           | `GEMINI_API_KEY` or `GOOGLE_API_KEY` | gemini-1.5-pro  |
| Ollama    | any locally hosted model | `OLLAMA_HOST` (endpoint URL)         | configurable    |

## What LLM Enhances

When an LLM provider is configured, Dorgu uses it to improve several areas of analysis:

* **Deeper code analysis** -- understands business logic, not just file structure
* **Framework-specific recommendations** -- tailored settings for Express, Spring Boot, Django, etc.
* **Resource sizing** -- more accurate CPU/memory recommendations based on application patterns
* **Security best practices** -- context-aware security recommendations beyond generic defaults

***

## OpenAI

### Setup

```bash theme={null}
# Option 1: Environment variable
export OPENAI_API_KEY="sk-proj-..."

# Option 2: Global config
dorgu config set llm.provider openai
dorgu config set llm.api_key "sk-proj-..."
```

### Usage

```bash theme={null}
dorgu generate . --llm-provider openai
```

### Model Override

```bash theme={null}
dorgu config set llm.model gpt-3.5-turbo
```

***

## Anthropic

### Setup

```bash theme={null}
# Option 1: Environment variable
export ANTHROPIC_API_KEY="sk-ant-..."

# Option 2: Global config
dorgu config set llm.provider anthropic
dorgu config set llm.api_key "sk-ant-..."
```

### Usage

```bash theme={null}
dorgu generate . --llm-provider anthropic
```

***

## Gemini

### Setup

```bash theme={null}
# Option 1: Environment variable (either key works)
export GEMINI_API_KEY="AIza..."
# or
export GOOGLE_API_KEY="AIza..."

# Option 2: Global config
dorgu config set llm.provider gemini
dorgu config set llm.api_key "AIza..."
```

### Usage

```bash theme={null}
dorgu generate . --llm-provider gemini
```

***

## Ollama

Ollama lets you run LLMs locally without an API key. You need a running Ollama instance with at least one model pulled.

### Prerequisites

```bash theme={null}
# Install Ollama (see https://ollama.ai)
# Pull a model
ollama pull llama3
```

### Setup

```bash theme={null}
# Option 1: Environment variable for custom endpoint
export OLLAMA_HOST="http://localhost:11434"

# Option 2: Global config
dorgu config set llm.provider ollama
dorgu config set llm.model llama3
```

### Usage

```bash theme={null}
dorgu generate . --llm-provider ollama
```

<Note>
  Ollama does not require an API key. The `OLLAMA_HOST` variable is only needed if your Ollama instance runs on a non-default address.
</Note>

***

## API Key Resolution Order

When Dorgu needs an API key, it checks the following sources in order:

1. **Environment variable** -- `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GEMINI_API_KEY`, or `GOOGLE_API_KEY`
2. **Global config** -- `llm.api_key` in `~/.config/dorgu/config.yaml`
3. **Interactive prompt** -- if running in a terminal and no key is found, Dorgu prompts you to enter one

## Setting Provider in `.dorgu.yaml`

You can also set the LLM provider at the workspace or app level:

```yaml theme={null}
# In .dorgu.yaml
llm:
  provider: "openai"
  model: "gpt-4"
```

This is useful when different projects require different providers or models.