# Hume Provider

Hume provides expressive text-to-speech synthesis. It does not offer
language, embedding, image, or transcription models — `LanguageModel`,
`EmbeddingModel`, `ImageModel`, and `TranscriptionModel` all return errors.

## Setup

### Installation

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

### Configuration

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

`hume.Config` also accepts `BaseURL` (default `https://api.hume.ai`).

### Get API Key

```bash
export HUME_API_KEY=...
```

## Speech Synthesis

Hume has no selectable model ID — `SpeechModel`'s `modelID` argument is
accepted for interface parity but ignored, matching the TypeScript SDK's
`speech()` factory (which takes no model ID parameter):

```go
speechModel, err := provider.SpeechModel("")
if err != nil {
    log.Fatal(err)
}

result, err := ai.GenerateSpeech(ctx, ai.GenerateSpeechOptions{
    Model: speechModel,
    Text:  "Hello from Hume.",
})
```

### Multi-Utterance Context

Pass a list of utterances (each with its own voice, speed, description,
and trailing silence), or continue a previous generation by ID, through
`providerOptions.hume.context`:

```go
result, err := ai.GenerateSpeech(ctx, ai.GenerateSpeechOptions{
    Model: speechModel,
    Text:  "", // ignored when context.utterances is set
    ProviderOptions: map[string]interface{}{
        "hume": hume.SpeechModelOptions{
            Context: &hume.Context{
                Utterances: []hume.Utterance{
                    {
                        Text:        "Welcome to the show.",
                        Description: "warm, welcoming tone",
                        Voice:       &hume.UtteranceVoice{Name: "Ava Song", Provider: "HUME_AI"},
                    },
                },
            },
        },
    },
})
```

To continue a previous generation instead of specifying utterances, set
`Context.GenerationID` (leave `Utterances` empty).

## Workflow Serialization

Hume speech models can cross a workflow boundary with
`provider.SerializeSpeechModel` / `DeserializeSpeechModel`. See
[Provider Serialization](https://goaisdk.com/docs/agents/workflow-agent.md#provider-serialization)
for the mechanism.

## See Also

- [API Reference: GenerateSpeech](https://goaisdk.com/docs/reference/ai/generate-speech.md)
- [Hume Documentation](https://dev.hume.ai)