# MiniMax Provider

MiniMax provides chat language models over an Anthropic-compatible
endpoint, plus its own video generation API. Chat requests delegate to the
shared Anthropic language model implementation (`Provider.Name()` is
`"minimax"`), so `providerOptions.minimax` and `providerOptions.anthropic`
are both read on every call.

## Setup

### Installation

```go
import (
    "github.com/digitallysavvy/go-ai/pkg/ai"
    "github.com/digitallysavvy/go-ai/pkg/providers/minimax"
)
```

### Configuration

```go
provider := minimax.New(minimax.Config{
    APIKey: os.Getenv("MINIMAX_API_KEY"),
})

model, err := provider.LanguageModel(minimax.ModelM3)
```

`minimax.Config` also accepts `BaseURL` (default
`https://api.minimax.io/anthropic/v1`), `VideoBaseURL` (default
`https://api.minimax.io`), and `Headers`. Video-request credentials are
resolved fresh on every call, so rotating `MINIMAX_API_KEY` after `New()`
still takes effect on the next request.

### Get API Key

```bash
export MINIMAX_API_KEY=...
```

## Available Chat Models

| Model ID | Go constant |
|----------|-------------|
| `minimax-m3` | `minimax.ModelM3` |
| `minimax-m2.7` | `minimax.ModelM27` |
| `minimax-m2.7-highspeed` | `minimax.ModelM27HighSpeed` |
| `minimax-m2.5` | `minimax.ModelM25` |
| `minimax-m2.5-highspeed` | `minimax.ModelM25HighSpeed` |
| `minimax-m2.1` | `minimax.ModelM21` |
| `minimax-m2.1-highspeed` | `minimax.ModelM21HighSpeed` |
| `minimax-m2` | `minimax.ModelM2` |

`ChatModel` is an alias for `LanguageModel`. `EmbeddingModel`,
`ImageModel`, `SpeechModel`, `TranscriptionModel`, and `RerankingModel`
all return errors — MiniMax only supports chat and video here.

## Basic Usage

```go
result, err := ai.GenerateText(ctx, ai.GenerateTextOptions{
    Model:  model,
    Prompt: "Explain the MiniMax M3 model in one paragraph",
})
```

## Video Generation

```go
videoModel, err := provider.VideoModel(string(minimax.ModelH3))
if err != nil {
    log.Fatal(err)
}

result, err := ai.GenerateVideo(ctx, ai.GenerateVideoOptions{
    Model:  videoModel,
    Prompt: ai.VideoPrompt{Text: "A neon-lit city street at night, rain reflections"},
})
```

Model IDs: `minimax.ModelH3` (`MiniMax-H3`), `minimax.ModelH3Max`
(`MiniMax-H3-Max`). Default resolution is `"2K"` and default aspect ratio
is `"16:9"`. `VideoModel` also implements `provider.VideoModelStarter` /
`VideoModelStatusChecker`, so it works with `ai.ExperimentalStartVideo` /
`ai.ExperimentalGetVideoStatus`.

## Provider Options

Since chat delegates to the Anthropic language model, MiniMax accepts the
same per-call `providerOptions.anthropic` fields (thinking, context
management, etc. — see the [Anthropic provider](https://goaisdk.com/docs/providers/anthropic.md)) under
either the `"minimax"` or `"anthropic"` key.

## See Also

- [API Reference: GenerateText](https://goaisdk.com/docs/reference/ai/generate-text.md)
- [Anthropic Provider](https://goaisdk.com/docs/providers/anthropic.md) - MiniMax chat shares this implementation
- [MiniMax Documentation](https://www.minimax.io/platform_overview)