talk/internal/provider/elevenlabs_test.go

package provider

import (
	"context"
	"encoding/json"
	"io"
	"net/http"
	"net/http/httptest"
	"strings"
	"sync"
	"testing"
)

// recordedRequest records a request seen by the fake server.
type recordedRequest struct {
	method string
	path   string
	header http.Header
	body   []byte
}

// fakeTTSServer is the httptest fake of a cloud provider: responds with
// status + body and records the (single) request it saw.
func fakeTTSServer(t *testing.T, status int, body []byte) (*httptest.Server, *recordedRequest) {
	t.Helper()
	rec := &recordedRequest{}
	var mu sync.Mutex
	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
		body2, err := io.ReadAll(r.Body)
		if err != nil {
			t.Errorf("fake server: reading body: %v", err)
		}
		mu.Lock()
		rec.method, rec.path, rec.header, rec.body = r.Method, r.URL.Path, r.Header.Clone(), body2
		mu.Unlock()
		w.WriteHeader(status)
		w.Write(body)
	}))
	t.Cleanup(ts.Close)
	return ts, rec
}

func envMap(kv map[string]string) func(string) string {
	return func(k string) string { return kv[k] }
}

// fakeMP3 is an arbitrary byte sequence; the providers return response
// bytes unchanged, the audio sink decodes (its own tests).
var fakeMP3 = []byte("\xff\xfb\x90\x00" + "audio-bytes-from-eleven")

// TestElevenLabsSynthesize200: 200 with bytes → bytes come back unchanged;
// method, path (incl. voice ID), header (xi-api-key) and the JSON body
// (model_id, voice_settings) match the API contract.
func TestElevenLabsSynthesize200(t *testing.T) {
	ts, rec := fakeTTSServer(t, 200, fakeMP3)
	p := NewElevenLabs(ChainConfig{
		Getenv: envMap(map[string]string{
			EnvElevenLabsKey:      "testkey",
			EnvElevenLabsEndpoint: ts.URL,
		}),
	})

	data, err := p.Synthesize(context.Background(), "Hello world", "de")
	if err != nil {
		t.Fatalf("Synthesize: %v", err)
	}
	if string(data) != string(fakeMP3) {
		t.Errorf("Synthesize = %q, want fake bytes %q", data, fakeMP3)
	}
	if rec.method != http.MethodPost {
		t.Errorf("method = %s, want POST", rec.method)
	}
	if want := "/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM"; rec.path != want {
		t.Errorf("path = %q, want %q (neutral repo default \"Rachel\")", rec.path, want)
	}
	if got := rec.header.Get("xi-api-key"); got != "testkey" {
		t.Errorf("xi-api-key = %q, want %q", got, "testkey")
	}

	var body struct {
		Text          string `json:"text"`
		ModelID       string `json:"model_id"`
		VoiceSettings struct {
			Stability       float64 `json:"stability"`
			SimilarityBoost float64 `json:"similarity_boost"`
		} `json:"voice_settings"`
	}
	if err := json.Unmarshal(rec.body, &body); err != nil {
		t.Fatalf("request body is not JSON: %v (%s)", err, rec.body)
	}
	if body.Text != "Hello world" {
		t.Errorf("text = %q", body.Text)
	}
	if body.ModelID != "eleven_multilingual_v2" {
		t.Errorf("model_id = %q, want eleven_multilingual_v2", body.ModelID)
	}
	if body.VoiceSettings.Stability != 0.5 || body.VoiceSettings.SimilarityBoost != 0.75 {
		t.Errorf("voice_settings = %+v, want stability 0.5 / similarity_boost 0.75", body.VoiceSettings)
	}
}

// TestElevenLabsHTTPError: 500 and 401 are clear errors naming the status
// and a body snippet (the chain falls back on them).
func TestElevenLabsHTTPError(t *testing.T) {
	cases := []struct {
		status int
		body   string
		want   string
	}{
		{500, "internal boom", "500"},
		{401, `{"detail":{"status":"invalid_api_key"}}`, "401"},
	}
	for _, tc := range cases {
		t.Run(http.StatusText(tc.status), func(t *testing.T) {
			ts, _ := fakeTTSServer(t, tc.status, []byte(tc.body))
			p := NewElevenLabs(ChainConfig{Getenv: envMap(map[string]string{
				EnvElevenLabsKey:      "testkey",
				EnvElevenLabsEndpoint: ts.URL,
			})})
			_, err := p.Synthesize(context.Background(), "Hello", "de")
			if err == nil || !strings.Contains(err.Error(), tc.want) {
				t.Fatalf("Synthesize error = %v, want HTTP %s", err, tc.want)
			}
			if !strings.Contains(err.Error(), "elevenlabs") {
				t.Errorf("error should name the provider: %v", err)
			}
		})
	}
}

// TestElevenLabsAvailable: not available without a key, available with
// one; without any network call.
func TestElevenLabsAvailable(t *testing.T) {
	if NewElevenLabs(ChainConfig{Getenv: envMap(nil)}).Available(context.Background()) {
		t.Error("Available without key = true, want false")
	}
	p := NewElevenLabs(ChainConfig{Getenv: envMap(map[string]string{EnvElevenLabsKey: "k"})})
	if !p.Available(context.Background()) {
		t.Error("Available with key = false, want true")
	}
}

// TestElevenLabsVoicePrecedence: the --voice pin beats the config mapping,
// the mapping beats the repo default (observable in the request path).
func TestElevenLabsVoicePrecedence(t *testing.T) {
	mk := func(t *testing.T, voices map[string]string, pin string) (*ElevenLabs, *recordedRequest) {
		t.Helper()
		ts, rec := fakeTTSServer(t, 200, fakeMP3)
		p := NewElevenLabs(ChainConfig{
			Getenv: envMap(map[string]string{
				EnvElevenLabsKey:      "k",
				EnvElevenLabsEndpoint: ts.URL,
			}),
			Voices: map[string]map[string]string{"elevenlabs": voices},
		})
		if pin != "" {
			p.PinVoice(pin)
		}
		return p, rec
	}

	t.Run("config mapping beats default", func(t *testing.T) {
		p, rec := mk(t, map[string]string{"de": "FromConfig"}, "")
		if _, err := p.Synthesize(context.Background(), "Hello", "de"); err != nil {
			t.Fatalf("Synthesize: %v", err)
		}
		if want := "/v1/text-to-speech/FromConfig"; rec.path != want {
			t.Errorf("path = %q, want %q", rec.path, want)
		}
	})
	t.Run("pin (--voice) beats config mapping", func(t *testing.T) {
		p, rec := mk(t, map[string]string{"de": "FromConfig"}, "Pinned")
		if _, err := p.Synthesize(context.Background(), "Hello", "de"); err != nil {
			t.Fatalf("Synthesize: %v", err)
		}
		if want := "/v1/text-to-speech/Pinned"; rec.path != want {
			t.Errorf("path = %q, want %q", rec.path, want)
		}
	})
	t.Run("unknown language falls back to en default", func(t *testing.T) {
		p, rec := mk(t, nil, "")
		if _, err := p.Synthesize(context.Background(), "Hello", "fr"); err != nil {
			t.Fatalf("Synthesize: %v", err)
		}
		if want := "/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM"; rec.path != want {
			t.Errorf("path = %q, want en default %q", rec.path, want)
		}
	})
}

// TestElevenLabsEmptyText is a clear error without a request.
func TestElevenLabsEmptyText(t *testing.T) {
	ts, rec := fakeTTSServer(t, 200, fakeMP3)
	p := NewElevenLabs(ChainConfig{Getenv: envMap(map[string]string{
		EnvElevenLabsKey:      "k",
		EnvElevenLabsEndpoint: ts.URL,
	})})
	if _, err := p.Synthesize(context.Background(), "", "de"); err == nil {
		t.Error("Synthesize(\"\") should fail")
	}
	if rec.method != "" {
		t.Error("no request must be sent for empty text")
	}
}