# multillm-openrouter

OpenRouter provider for [multillm](../multillm).

OpenRouter provides access to many models through a single API.

## Installation

```bash
pip install multillm-openrouter
```

## Usage

```python
import asyncio
import multillm

async def main():
    client = multillm.Client()

    # Simple query
    answer = await client.single("openrouter/anthropic/claude-3-haiku", "Hello!")
    print(answer)

    # Chat completion
    response = await client.chat_complete("openrouter/meta-llama/llama-3-70b-instruct", [
        {"role": "user", "content": "What is Python?"}
    ])
    print(response.text)

    # Streaming
    async for chunk in client.chat_complete_stream("openrouter/openai/gpt-4o", messages):
        print(chunk, end="")

asyncio.run(main())
```

## Configuration

Choose one of the following methods (priority: direct config > env var > config file):

**Option 1: Config file** (recommended)

Create `~/.config/multillm/providers/openrouter.json`:

```json
{
  "api_key": "sk-or-...",
  "base_url": "https://openrouter.ai/api/v1"
}
```

Set permissions:
```bash
chmod 600 ~/.config/multillm/providers/openrouter.json
```

**Option 2: Environment variable**

```bash
export OPENROUTER_API_KEY=sk-or-...
```

**Option 3: Direct config**

```python
client = multillm.Client({
    "openrouter": {"api_key": "sk-or-..."}
})
```

### Config File Fields

| Field | Required | Description |
|-------|----------|-------------|
| `api_key` | Yes | Your OpenRouter API key |
| `base_url` | No | API base URL (defaults to `https://openrouter.ai/api/v1`) |

## Models

Use OpenRouter model IDs after the provider prefix. Models follow the format `openrouter/<provider>/<model>`:

- `openrouter/anthropic/claude-3-haiku`
- `openrouter/openai/gpt-4o`
- `openrouter/meta-llama/llama-3-70b-instruct`
- `openrouter/google/gemini-pro`

See [OpenRouter models](https://openrouter.ai/models) for the full list.