Maven/internal/stt/stt_test.go

package stt

import (
	"context"
	"testing"

	"github.com/kami/maven/internal/audio"
	"github.com/kami/maven/internal/worker"
)

// fakeTranscriber — a worker.Transcriber that records the call.
type fakeTranscriber struct {
	got audio.Audio
}

func (f *fakeTranscriber) Transcribe(_ context.Context, req worker.TranscribeReq) (worker.TranscribeResp, error) {
	f.got = req.Audio
	return worker.TranscribeResp{Text: "hello world", Confidence: 0.7}, nil
}

func TestStubReturnsDeterministicPhrase(t *testing.T) {
	t.Parallel()
	s := NewStub()
	a := audio.Audio{Format: audio.PCM16kMono, Bytes: []byte("utterance-A")}
	text, conf, err := s.Transcribe(context.Background(), a)
	if err != nil {
		t.Fatalf("Stub.Transcribe: %v", err)
	}
	if conf != 1.0 {
		t.Fatalf("Stub confidence should be 1.0, got %v", conf)
	}
	// same input ⇒ same output (deterministic; tests can rely on this).
	text2, _, _ := s.Transcribe(context.Background(), a)
	if text != text2 {
		t.Fatalf("Stub should be deterministic: %q vs %q", text, text2)
	}
	// phrases come from the stubPhrases list.
	found := false
	for _, p := range stubPhrases {
		if p == text {
			found = true
			break
		}
	}
	if !found {
		t.Fatalf("Stub phrase %q not in stubPhrases", text)
	}
}

func TestStubEmptyAudioReturnsFirstPhrase(t *testing.T) {
	t.Parallel()
	s := NewStub()
	text, _, err := s.Transcribe(context.Background(), audio.Audio{Format: audio.PCM16kMono})
	if err != nil {
		t.Fatalf("Stub.Transcribe: %v", err)
	}
	if text != stubPhrases[0] {
		t.Fatalf("empty audio should return first phrase %q, got %q", stubPhrases[0], text)
	}
}

func TestStubDifferentAudioMayPickDifferentPhrase(t *testing.T) {
	t.Parallel()
	s := NewStub()
	// try enough variants to land at least two different phrases.
	seen := map[string]struct{}{}
	for i := 0; i < 200; i++ {
		b := make([]byte, 64)
		for j := range b {
			b[j] = byte(i + j)
		}
		text, _, _ := s.Transcribe(context.Background(), audio.Audio{Format: audio.PCM16kMono, Bytes: b})
		seen[text] = struct{}{}
		if len(seen) >= 2 {
			return
		}
	}
	t.Fatalf("expected ≥2 distinct phrases across varied audio, got %d", len(seen))
}

func TestRemoteForwardsToClient(t *testing.T) {
	t.Parallel()
	// stand up a worker.Server with the fakeTranscriber behind it.
	srv := worker.NewServer(t.TempDir()+"/stt.sock", &fakeTranscriber{})
	if err := srv.Listen(); err != nil {
		t.Fatalf("listen: %v", err)
	}
	defer srv.Close()
	go srv.Serve()

	c := worker.Dial(srv.Path())
	defer c.Close()
	r := NewRemote(c, "ru")

	in := audio.Audio{Format: audio.PCM16kMono, Bytes: []byte("payload")}
	text, conf, err := r.Transcribe(context.Background(), in)
	if err != nil {
		t.Fatalf("Remote.Transcribe: %v", err)
	}
	if text != "hello world" {
		t.Fatalf("Text: %q, want %q", text, "hello world")
	}
	if conf != 0.7 {
		t.Fatalf("Confidence: %v, want 0.7", conf)
	}
}

func TestRemoteErrorWraps(t *testing.T) {
	t.Parallel()
	// a server with no transcriber ⇒ ErrUnknownMethod on Transcribe.
	srv := worker.NewSynthesizerServer(t.TempDir()+"/stt.sock", nil)
	if err := srv.Listen(); err != nil {
		t.Fatalf("listen: %v", err)
	}
	defer srv.Close()
	go srv.Serve()

	c := worker.Dial(srv.Path())
	defer c.Close()
	r := NewRemote(c, "ru")
	_, _, err := r.Transcribe(context.Background(), audio.Audio{Format: audio.PCM16kMono, Bytes: []byte("x")})
	if err == nil {
		t.Fatalf("want error, got nil")
	}
}