talk/internal/provider/elevenlabs.go

package provider

import (
	"context"
	"encoding/json"
	"fmt"
	"net/http"
	"strings"
)

// ElevenLabs constants (public contract: the env names are part of the
// fleet documentation, the endpoint default part of the configurability).
const (
	// DefaultEndpointElevenLabs is the production API endpoint.
	DefaultEndpointElevenLabs = "https://api.elevenlabs.io"

	// EnvElevenLabsKey provides the API key (spec: keys come exclusively
	// from the environment, never from config.toml).
	EnvElevenLabsKey = "TALK_ELEVENLABS_KEY"

	// EnvElevenLabsEndpoint overrides the endpoint (tests/corporate
	// servers; beats config.toml, see the voices.go/chain.go docs).
	EnvElevenLabsEndpoint = "TALK_ELEVENLABS_ENDPOINT"

	elevenLabsModel = "eleven_multilingual_v2" // multilingual v2
)

// ElevenLabs is the premium tier of the chain. Synthesize performs POST
// <endpoint>/v1/text-to-speech/<voiceID> with header `xi-api-key` and body
// {text, model_id: "eleven_multilingual_v2", voice_settings:
// {stability: 0.5, similarity_boost: 0.75}} and returns the response bytes
// unchanged. Return format: MP3 (documented contract of the audio package).
type ElevenLabs struct {
	// Endpoint is the base URL (without path); empty is replaced by the
	// default at construction.
	Endpoint string

	// APIKey is the ElevenLabs key (from TALK_ELEVENLABS_KEY).
	APIKey string

	// Voices is the config mapping voice.elevenlabs.<lang> = id; empty =
	// neutral repo defaults (see voices.go).
	Voices map[string]string

	// Client is the HTTP client; nil = shared client with timeout
	// (cloudTimeout).
	Client *http.Client

	pinned string // --voice pin (VoicePinner)
}

// NewElevenLabs builds the provider from the chain configuration: endpoint
// from TALK_ELEVENLABS_ENDPOINT > [providers.elevenlabs].endpoint > default;
// key exclusively from TALK_ELEVENLABS_KEY.
func NewElevenLabs(cfg ChainConfig) *ElevenLabs {
	return &ElevenLabs{
		Endpoint: endpointFrom(cfg, "elevenlabs", EnvElevenLabsEndpoint, DefaultEndpointElevenLabs),
		APIKey:   cfg.Getenv(EnvElevenLabsKey),
		Voices:   cfg.Voices["elevenlabs"],
	}
}

// Name implements Provider.
func (*ElevenLabs) Name() string { return "elevenlabs" }

// PinVoice implements VoicePinner (CLI flag --voice).
func (p *ElevenLabs) PinVoice(id string) { p.pinned = id }

// Available implements Provider: key and endpoint must be set (deliberately
// without a network call; Available must stay cheap. Whether the key is
// valid is determined by Synthesize, and the chain falls back).
func (p *ElevenLabs) Available(ctx context.Context) bool {
	return p.APIKey != "" && p.Endpoint != ""
}

// elevenLabsRequest is the request body (literal values from the API
// contract).
type elevenLabsRequest struct {
	Text          string                  `json:"text"`
	ModelID       string                  `json:"model_id"`
	VoiceSettings elevenLabsVoiceSettings `json:"voice_settings"`
}

type elevenLabsVoiceSettings struct {
	Stability       float64 `json:"stability"`
	SimilarityBoost float64 `json:"similarity_boost"`
}

// Synthesize implements Provider. Return format: MP3.
func (p *ElevenLabs) Synthesize(ctx context.Context, text, lang string) ([]byte, error) {
	if text == "" {
		return nil, fmt.Errorf("elevenlabs: empty text")
	}
	voice := resolveVoice("elevenlabs", lang, p.pinned, p.Voices)
	if voice == "" {
		return nil, fmt.Errorf("elevenlabs: no voice configured for language %q", lang)
	}
	body, err := json.Marshal(elevenLabsRequest{
		Text:          text,
		ModelID:       elevenLabsModel,
		VoiceSettings: elevenLabsVoiceSettings{Stability: 0.5, SimilarityBoost: 0.75},
	})
	if err != nil {
		return nil, fmt.Errorf("elevenlabs: building request body: %w", err)
	}
	url := strings.TrimRight(p.Endpoint, "/") + "/v1/text-to-speech/" + voice
	return postForBytes(ctx, p.Client, "elevenlabs", url,
		map[string]string{
			"xi-api-key": p.APIKey,
			"Accept":     "audio/mpeg",
		}, body)
}