Add Chromium to websocket proxy.

Author: goodov

.gitignore

@@ -24,3 +24,5 @@ models/*
 third_party/whisper.cpp
 .env
 dist
+
+/.vscode

cmd/crtowebsocket/README.md

@@ -0,0 +1,36 @@
+# Chromium to WebSocket
+
+This is a simple HTTP server that listens for WebSpeech connections from a
+Chromium browser, converts audio data to WAV format, streams it to a WebSocket
+client, receives text data from the WebSocket client, and sends it back to the
+Chromium browser.
+
+## Usage
+
+```bash
+go run .
+```
+
+## Building
+
+```bash
+go build -o crtowebsocket .
+```
+
+## Running
+
+```bash
+./crtowebocket
+```
+
+## Building for release
+
+```bash
+go build -o crtowebsocket .
+```
+
+## Running for release
+
+```bash
+./crtowebsocket
+```

cmd/crtowebsocket/main.go

@@ -0,0 +1,91 @@
+package main
+
+import (
+	"net/http"
+	"os"
+	"strconv"
+	"time"
+
+	"github.com/brave-experiments/go-stt/cr_api_websocket_proxy"
+	"github.com/rs/zerolog"
+	"github.com/rs/zerolog/log"
+	"github.com/urfave/cli/v2"
+)
+
+// Configuration for the remote WebSocket STT service
+const (
+	version              = "1"
+	defaultListenAddress = "127.0.0.1:8090"
+	defaultWebsocketURL  = "ws://127.0.0.1:8080/api-speech-wss/"
+)
+
+func main() {
+	zerolog.SetGlobalLevel(zerolog.InfoLevel)
+	log.Logger = log.Output(
+		zerolog.ConsoleWriter{
+			Out:     os.Stderr,
+			NoColor: true,
+		},
+	)
+	zerolog.CallerMarshalFunc = func(pc uintptr, file string, line int) string {
+		short := file
+		for i := len(file) - 1; i > 0; i-- {
+			if file[i] == '/' {
+				short = file[i+1:]
+				break
+			}
+		}
+		file = short
+		return file + ":" + strconv.Itoa(line)
+	}
+
+	zerolog.SetGlobalLevel(zerolog.DebugLevel)
+	log.Logger = log.With().Caller().Logger()
+
+	app := cli.NewApp()
+	app.Name = "Chromium WebSpeech API Endpoint to WebSocket proxy"
+	app.Version = version
+	app.Flags = []cli.Flag{
+		&cli.StringFlag{
+			Name:  "listen-address",
+			Value: defaultListenAddress,
+		},
+		&cli.StringFlag{
+			Name:  "websocket-url",
+			Value: defaultWebsocketURL,
+		},
+		&cli.DurationFlag{
+			Name:  "timeout",
+			Value: 60 * time.Second,
+		},
+		&cli.BoolFlag{
+			Name:  "try-to-finalize-text",
+			Value: false,
+		},
+	}
+	app.Action = run
+
+	if err := app.Run(os.Args); err != nil {
+		log.Fatal().Err(err)
+	}
+}
+
+func run(c *cli.Context) error {
+	// Create a configuration struct
+	config := &cr_api_websocket_proxy.HandlerConfig{
+		WebsocketURL:      c.String("websocket-url"),
+		Timeout:           c.Duration("timeout"),
+		TryToFinalizeText: c.Bool("try-to-finalize-text"),
+	}
+
+	// Create a handler instance with the config
+	handler := cr_api_websocket_proxy.NewHandler(config)
+
+	// Register handlers that have access to the config
+	http.HandleFunc("/up", handler.HandleUpstreamRequest)
+	http.HandleFunc("/down", handler.HandleDownstreamRequest)
+
+	http.ListenAndServe(c.String("listen-address"), nil)
+
+	return nil
+}

cmd/crtowebsocket/pprof.go

@@ -0,0 +1,7 @@
+//go:build pprof
+
+package main
+
+import (
+	_ "net/http/pprof"
+)

cr_api_websocket_proxy/audio.go

@@ -0,0 +1,51 @@
+package cr_api_websocket_proxy
+
+import (
+	"fmt"
+	"net/http"
+
+	"azul3d.org/engine/audio"
+	"github.com/colega/zeropool"
+
+	// Add flac decoder for decoding incoming audio
+	_ "azul3d.org/engine/audio/flac"
+)
+
+const expectedSampleRate = 16000
+
+const (
+	samplesPerChunk = expectedSampleRate / 1000 * 20 // 20ms
+	bytesPerChunk   = samplesPerChunk * 2
+)
+
+var audioSamplesBufferPool = zeropool.New(
+	func() audio.Int16 {
+		return make(
+			audio.Int16,
+			samplesPerChunk,
+		)
+	},
+)
+
+var audioBytesBufferPool = zeropool.New(
+	func() []byte {
+		return make(
+			[]byte,
+			bytesPerChunk,
+		)
+	},
+)
+
+func NewAudioDecoder(req *http.Request) (audio.Decoder, error) {
+	dec, _, err := audio.NewDecoder(req.Body)
+	if err != nil {
+		return nil, err
+	}
+
+	// Ensure we're working with the correct sample rate
+	if dec.Config().SampleRate != expectedSampleRate {
+		return nil, fmt.Errorf("unexpected sample rate: %d", dec.Config().SampleRate)
+	}
+
+	return dec, nil
+}

cr_api_websocket_proxy/audio_test.go

@@ -0,0 +1,112 @@
+package cr_api_websocket_proxy
+
+import (
+	"bytes"
+	"io"
+	"net/http"
+	"os"
+	"testing"
+
+	"azul3d.org/engine/audio"
+)
+
+func TestAudioBufferPools(t *testing.T) {
+	t.Run("samples buffer pool", func(t *testing.T) {
+		samples := audioSamplesBufferPool.Get()
+		if samples == nil {
+			t.Error("expected non-nil samples buffer")
+		}
+		if got := len(samples); got != samplesPerChunk {
+			t.Errorf("samples buffer length = %v, want %v", got, samplesPerChunk)
+		}
+
+		audioSamplesBufferPool.Put(samples)
+	})
+
+	t.Run("bytes buffer pool", func(t *testing.T) {
+		buffer := audioBytesBufferPool.Get()
+		if buffer == nil {
+			t.Error("expected non-nil bytes buffer")
+		}
+		if got := len(buffer); got != bytesPerChunk {
+			t.Errorf("bytes buffer length = %v, want %v", got, bytesPerChunk)
+		}
+
+		audioBytesBufferPool.Put(buffer)
+	})
+}
+
+type mockBody struct {
+	*bytes.Buffer
+}
+
+func (m mockBody) Close() error {
+	return nil
+}
+
+func TestFlacDecoder_InvalidData(t *testing.T) {
+	t.Run("invalid audio data", func(t *testing.T) {
+		req := &http.Request{
+			Body: mockBody{bytes.NewBuffer([]byte("invalid audio data"))},
+		}
+
+		decoder, err := NewAudioDecoder(req)
+		if err == nil {
+			t.Error("expected error for invalid audio data")
+		}
+		if decoder != nil {
+			t.Error("expected nil decoder for invalid audio data")
+		}
+	})
+}
+
+func TestFlacDecoder_ValidData(t *testing.T) {
+	req := &http.Request{
+		Body: mockBody{bytes.NewBuffer(readTestFile(t, "testdata/16khz.flac"))},
+	}
+
+	decoder, err := NewAudioDecoder(req)
+	if err != nil {
+		t.Fatalf("failed to create decoder: %v", err)
+	}
+
+	// Verify decoder config
+	config := decoder.Config()
+	if config.SampleRate != expectedSampleRate {
+		t.Errorf("sample rate = %v, want %v", config.SampleRate, expectedSampleRate)
+	}
+
+	// Try reading some samples
+	samples := make(audio.Int16, 1024)
+	n, err := decoder.Read(samples)
+	if err != nil && err != io.EOF {
+		t.Errorf("failed to read samples: %v", err)
+	}
+	if n == 0 {
+		t.Error("expected to read some samples")
+	}
+}
+
+func TestFlacDecoder_InvalidSampleRate(t *testing.T) {
+	req := &http.Request{
+		Body: mockBody{bytes.NewBuffer(readTestFile(t, "testdata/8khz.flac"))},
+	}
+
+	decoder, err := NewAudioDecoder(req)
+	if decoder != nil {
+		t.Error("expected nil decoder for invalid sample rate")
+	}
+	if err == nil {
+		t.Error("expected error for invalid sample rate")
+	}
+}
+
+// Helper to read test file contents
+func readTestFile(t *testing.T, path string) []byte {
+	t.Helper()
+	data, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("failed to read test file %s: %v", path, err)
+	}
+	return data
+}