add deepgram streaming adapter with reconnection logic

This commit is contained in:
leonardotrapani
2026-02-01 01:47:12 +01:00
parent 5f614b19c5
commit e9499b0d6b
4 changed files with 855 additions and 2 deletions
+406
View File
@@ -0,0 +1,406 @@
package transcriber
import (
"context"
"encoding/json"
"fmt"
"log"
"net/http"
"net/url"
"sync"
"time"
"github.com/gorilla/websocket"
"github.com/leonardotrapani/hyprvoice/internal/language"
"github.com/leonardotrapani/hyprvoice/internal/provider"
)
// DeepgramAdapter implements StreamingAdapter for Deepgram real-time transcription
type DeepgramAdapter struct {
endpoint *provider.EndpointConfig
apiKey string
model string
language string
conn *websocket.Conn
resultsCh chan TranscriptionResult
mu sync.Mutex
ctx context.Context
cancel context.CancelFunc
wg sync.WaitGroup
started bool
// reconnection config
maxRetries int
retryDelays []time.Duration
}
// Deepgram WebSocket response types (incoming)
type deepgramWSResponse struct {
Type string `json:"type"`
Channel *deepgramChannel `json:"channel,omitempty"`
Metadata *deepgramMetadata `json:"metadata,omitempty"`
Error *deepgramError `json:"error,omitempty"`
ChannelIdx []int `json:"channel_index,omitempty"`
Duration float64 `json:"duration,omitempty"`
Start float64 `json:"start,omitempty"`
IsFinal bool `json:"is_final,omitempty"`
SpeechFinal bool `json:"speech_final,omitempty"`
}
type deepgramChannel struct {
Alternatives []deepgramAlternative `json:"alternatives,omitempty"`
}
type deepgramAlternative struct {
Transcript string `json:"transcript"`
Confidence float64 `json:"confidence"`
}
type deepgramMetadata struct {
RequestID string `json:"request_id"`
ModelInfo struct {
Name string `json:"name"`
Version string `json:"version"`
} `json:"model_info"`
}
type deepgramError struct {
Type string `json:"type"`
Message string `json:"message"`
Description string `json:"description,omitempty"`
}
// NewDeepgramAdapter creates a new streaming adapter for Deepgram
// endpoint: the WebSocket endpoint config (e.g., wss://api.deepgram.com, /v1/listen)
// apiKey: Deepgram API key
// model: model ID (e.g., "nova-3")
// lang: canonical language code (will be converted to provider format)
func NewDeepgramAdapter(endpoint *provider.EndpointConfig, apiKey, model, lang string) *DeepgramAdapter {
return &DeepgramAdapter{
endpoint: endpoint,
apiKey: apiKey,
model: model,
language: lang,
resultsCh: make(chan TranscriptionResult, 100),
maxRetries: 3,
retryDelays: defaultRetryDelays,
}
}
// Start initiates the WebSocket connection to Deepgram
func (a *DeepgramAdapter) Start(ctx context.Context, lang string) error {
a.mu.Lock()
defer a.mu.Unlock()
if a.started {
return fmt.Errorf("adapter already started")
}
// use lang param if provided, otherwise use constructor lang
if lang != "" {
a.language = lang
}
// create cancelable context
a.ctx, a.cancel = context.WithCancel(ctx)
// connect to WebSocket
if err := a.connectLocked(); err != nil {
return err
}
a.started = true
// start reader goroutine
a.wg.Add(1)
go a.readLoop()
log.Printf("deepgram: connected, model=%s, language=%s", a.model, a.language)
return nil
}
// connectLocked establishes WebSocket connection. Must be called with mu held.
func (a *DeepgramAdapter) connectLocked() error {
wsURL, err := a.buildURL()
if err != nil {
return fmt.Errorf("build websocket url: %w", err)
}
headers := http.Header{}
headers.Set("Authorization", "Token "+a.apiKey)
log.Printf("deepgram: connecting to %s", wsURL)
conn, resp, err := websocket.DefaultDialer.DialContext(a.ctx, wsURL, headers)
if err != nil {
if resp != nil {
log.Printf("deepgram: dial failed with status %d", resp.StatusCode)
}
return fmt.Errorf("websocket dial: %w", err)
}
a.conn = conn
return nil
}
// reconnect attempts to re-establish the WebSocket connection with exponential backoff.
// Returns true if reconnection succeeded.
func (a *DeepgramAdapter) reconnect() bool {
for attempt := 0; attempt < a.maxRetries; attempt++ {
// check if context cancelled
select {
case <-a.ctx.Done():
return false
default:
}
// wait before retry (skip wait on first attempt)
if attempt > 0 {
delay := a.retryDelays[attempt-1]
if attempt-1 >= len(a.retryDelays) {
delay = a.retryDelays[len(a.retryDelays)-1]
}
log.Printf("deepgram: reconnect attempt %d/%d after %v", attempt+1, a.maxRetries, delay)
select {
case <-a.ctx.Done():
return false
case <-time.After(delay):
}
} else {
log.Printf("deepgram: reconnect attempt %d/%d", attempt+1, a.maxRetries)
}
a.mu.Lock()
// close old connection if exists
if a.conn != nil {
a.conn.Close()
a.conn = nil
}
err := a.connectLocked()
a.mu.Unlock()
if err == nil {
log.Printf("deepgram: reconnected successfully")
// notify caller of brief interruption
select {
case a.resultsCh <- TranscriptionResult{Error: fmt.Errorf("connection interrupted, reconnected"), IsFinal: false}:
default:
}
return true
}
log.Printf("deepgram: reconnect failed: %v", err)
}
return false
}
// buildURL constructs the WebSocket URL with query parameters
func (a *DeepgramAdapter) buildURL() (string, error) {
// parse base URL and path
baseURL := a.endpoint.BaseURL + a.endpoint.Path
u, err := url.Parse(baseURL)
if err != nil {
return "", fmt.Errorf("parse base url: %w", err)
}
// add query parameters
q := u.Query()
q.Set("model", a.model)
q.Set("encoding", "linear16") // 16-bit linear PCM
q.Set("sample_rate", "16000") // 16kHz
q.Set("channels", "1") // mono
// enable interim results
q.Set("interim_results", "true")
// enable smart formatting for better output
q.Set("smart_format", "true")
// enable punctuation
q.Set("punctuate", "true")
// add language if specified
providerLang := language.ToProviderFormat(a.language, "deepgram")
if providerLang != "" {
q.Set("language", providerLang)
}
u.RawQuery = q.Encode()
return u.String(), nil
}
// readLoop reads messages from the WebSocket and sends results to the channel
func (a *DeepgramAdapter) readLoop() {
defer a.wg.Done()
defer close(a.resultsCh)
for {
select {
case <-a.ctx.Done():
return
default:
}
a.mu.Lock()
conn := a.conn
a.mu.Unlock()
if conn == nil {
// no connection, try to reconnect
if !a.reconnect() {
a.resultsCh <- TranscriptionResult{Error: fmt.Errorf("connection lost, reconnection failed after %d attempts", a.maxRetries)}
return
}
continue
}
_, message, err := conn.ReadMessage()
if err != nil {
// check if context was cancelled (normal shutdown)
select {
case <-a.ctx.Done():
return
default:
}
// attempt reconnection
log.Printf("deepgram: read error: %v, attempting reconnection", err)
if !a.reconnect() {
a.resultsCh <- TranscriptionResult{Error: fmt.Errorf("websocket read: %w, reconnection failed", err)}
return
}
continue
}
// parse message
var resp deepgramWSResponse
if err := json.Unmarshal(message, &resp); err != nil {
log.Printf("deepgram: parse error: %v", err)
continue
}
// handle different message types
switch resp.Type {
case "Metadata":
if resp.Metadata != nil {
log.Printf("deepgram: session started, request_id=%s, model=%s",
resp.Metadata.RequestID, resp.Metadata.ModelInfo.Name)
}
case "Results":
// transcription result
if resp.Channel != nil && len(resp.Channel.Alternatives) > 0 {
transcript := resp.Channel.Alternatives[0].Transcript
if transcript != "" {
isFinal := resp.IsFinal || resp.SpeechFinal
if isFinal {
log.Printf("deepgram: final: %q", transcript)
}
a.resultsCh <- TranscriptionResult{Text: transcript, IsFinal: isFinal}
}
}
case "Error":
if resp.Error != nil {
errMsg := resp.Error.Message
if resp.Error.Description != "" {
errMsg = fmt.Sprintf("%s: %s", errMsg, resp.Error.Description)
}
log.Printf("deepgram: error: %s", errMsg)
a.resultsCh <- TranscriptionResult{Error: fmt.Errorf("deepgram: %s", errMsg)}
}
case "UtteranceEnd":
log.Printf("deepgram: utterance end detected")
case "SpeechStarted":
log.Printf("deepgram: speech started")
default:
log.Printf("deepgram: unknown message type: %s", resp.Type)
}
}
}
// SendChunk sends audio data to the WebSocket
// Deepgram expects raw binary audio data, not base64 encoded
func (a *DeepgramAdapter) SendChunk(audio []byte) error {
a.mu.Lock()
if !a.started {
a.mu.Unlock()
return fmt.Errorf("adapter not started")
}
conn := a.conn
a.mu.Unlock()
// check context
select {
case <-a.ctx.Done():
return a.ctx.Err()
default:
}
if conn == nil {
return fmt.Errorf("no connection")
}
// send raw binary audio (not base64)
a.mu.Lock()
err := a.conn.WriteMessage(websocket.BinaryMessage, audio)
a.mu.Unlock()
if err != nil {
// attempt reconnection
log.Printf("deepgram: write error: %v, attempting reconnection", err)
if a.reconnect() {
// retry the chunk after reconnection
a.mu.Lock()
err = a.conn.WriteMessage(websocket.BinaryMessage, audio)
a.mu.Unlock()
if err == nil {
return nil
}
}
return fmt.Errorf("websocket write: %w", err)
}
return nil
}
// Results returns the channel for receiving transcription results
func (a *DeepgramAdapter) Results() <-chan TranscriptionResult {
return a.resultsCh
}
// Close gracefully closes the WebSocket connection
func (a *DeepgramAdapter) Close() error {
a.mu.Lock()
if !a.started {
a.mu.Unlock()
return nil
}
// cancel context first to signal reader to stop
if a.cancel != nil {
a.cancel()
}
// get conn ref while holding lock
conn := a.conn
a.started = false
a.mu.Unlock()
// close websocket outside of lock (readLoop may be blocked on read)
if conn != nil {
// send close frame (best effort)
_ = conn.WriteMessage(websocket.CloseMessage, websocket.FormatCloseMessage(websocket.CloseNormalClosure, ""))
conn.Close()
}
// wait for reader to finish
a.wg.Wait()
log.Printf("deepgram: closed")
return nil
}
@@ -0,0 +1,434 @@
package transcriber
import (
"context"
"encoding/json"
"net/http"
"net/http/httptest"
"strings"
"sync"
"testing"
"time"
"github.com/gorilla/websocket"
"github.com/leonardotrapani/hyprvoice/internal/provider"
)
func TestDeepgramAdapter_ImplementsStreamingAdapter(t *testing.T) {
var _ StreamingAdapter = (*DeepgramAdapter)(nil)
}
func TestDeepgramAdapter_Creation(t *testing.T) {
endpoint := &provider.EndpointConfig{
BaseURL: "wss://api.deepgram.com",
Path: "/v1/listen",
}
adapter := NewDeepgramAdapter(endpoint, "test-api-key", "nova-3", "en")
if adapter.apiKey != "test-api-key" {
t.Errorf("apiKey = %q, want %q", adapter.apiKey, "test-api-key")
}
if adapter.model != "nova-3" {
t.Errorf("model = %q, want %q", adapter.model, "nova-3")
}
if adapter.language != "en" {
t.Errorf("language = %q, want %q", adapter.language, "en")
}
if adapter.maxRetries != 3 {
t.Errorf("maxRetries = %d, want %d", adapter.maxRetries, 3)
}
}
func TestDeepgramAdapter_BuildURL(t *testing.T) {
tests := []struct {
name string
model string
language string
wantURL []string // URL must contain all these substrings
}{
{
name: "english",
model: "nova-3",
language: "en",
wantURL: []string{"model=nova-3", "language=en-US", "encoding=linear16", "sample_rate=16000"},
},
{
name: "spanish",
model: "nova-2",
language: "es",
wantURL: []string{"model=nova-2", "language=es", "encoding=linear16"},
},
{
name: "auto-detect",
model: "nova-3",
language: "",
wantURL: []string{"model=nova-3", "encoding=linear16"},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
endpoint := &provider.EndpointConfig{
BaseURL: "wss://api.deepgram.com",
Path: "/v1/listen",
}
adapter := NewDeepgramAdapter(endpoint, "test-key", tt.model, tt.language)
url, err := adapter.buildURL()
if err != nil {
t.Fatalf("buildURL() error = %v", err)
}
for _, want := range tt.wantURL {
if !strings.Contains(url, want) {
t.Errorf("buildURL() = %q, want to contain %q", url, want)
}
}
})
}
}
func TestDeepgramAdapter_SendChunkNotStarted(t *testing.T) {
endpoint := &provider.EndpointConfig{
BaseURL: "wss://api.deepgram.com",
Path: "/v1/listen",
}
adapter := NewDeepgramAdapter(endpoint, "test-key", "nova-3", "en")
err := adapter.SendChunk([]byte("audio data"))
if err == nil {
t.Error("SendChunk() should return error when adapter not started")
}
if !strings.Contains(err.Error(), "not started") {
t.Errorf("error should mention 'not started', got: %v", err)
}
}
func TestDeepgramAdapter_CloseNotStarted(t *testing.T) {
endpoint := &provider.EndpointConfig{
BaseURL: "wss://api.deepgram.com",
Path: "/v1/listen",
}
adapter := NewDeepgramAdapter(endpoint, "test-key", "nova-3", "en")
// closing not-started adapter should not error
err := adapter.Close()
if err != nil {
t.Errorf("Close() error = %v, want nil", err)
}
}
// mockDeepgramServer creates a mock WebSocket server for testing
func mockDeepgramServer(t *testing.T, handler func(*websocket.Conn)) *httptest.Server {
upgrader := websocket.Upgrader{}
server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
// verify auth header
auth := r.Header.Get("Authorization")
if !strings.HasPrefix(auth, "Token ") {
http.Error(w, "Unauthorized", http.StatusUnauthorized)
return
}
conn, err := upgrader.Upgrade(w, r, nil)
if err != nil {
t.Logf("upgrade error: %v", err)
return
}
defer conn.Close()
handler(conn)
}))
return server
}
func TestDeepgramAdapter_StartAndClose(t *testing.T) {
server := mockDeepgramServer(t, func(conn *websocket.Conn) {
// send metadata response
metadata := deepgramWSResponse{
Type: "Metadata",
Metadata: &deepgramMetadata{
RequestID: "test-123",
},
}
metadata.Metadata.ModelInfo.Name = "nova-3"
if err := conn.WriteJSON(metadata); err != nil {
t.Logf("write metadata error: %v", err)
return
}
// wait for close
for {
_, _, err := conn.ReadMessage()
if err != nil {
break
}
}
})
defer server.Close()
// convert http URL to ws URL
wsURL := "ws" + strings.TrimPrefix(server.URL, "http")
endpoint := &provider.EndpointConfig{
BaseURL: wsURL,
Path: "",
}
adapter := NewDeepgramAdapter(endpoint, "test-api-key", "nova-3", "en")
ctx := context.Background()
if err := adapter.Start(ctx, ""); err != nil {
t.Fatalf("Start() error = %v", err)
}
// verify can't start twice
if err := adapter.Start(ctx, ""); err == nil {
t.Error("Start() should return error when already started")
}
// close
if err := adapter.Close(); err != nil {
t.Errorf("Close() error = %v", err)
}
}
func TestDeepgramAdapter_ReceivesResults(t *testing.T) {
var wg sync.WaitGroup
wg.Add(1)
server := mockDeepgramServer(t, func(conn *websocket.Conn) {
defer wg.Done()
// send metadata
metadata := deepgramWSResponse{Type: "Metadata", Metadata: &deepgramMetadata{RequestID: "test-123"}}
_ = conn.WriteJSON(metadata)
// send interim result
interim := deepgramWSResponse{
Type: "Results",
IsFinal: false,
Channel: &deepgramChannel{
Alternatives: []deepgramAlternative{{Transcript: "hello", Confidence: 0.95}},
},
}
_ = conn.WriteJSON(interim)
// send final result
final := deepgramWSResponse{
Type: "Results",
IsFinal: true,
Channel: &deepgramChannel{
Alternatives: []deepgramAlternative{{Transcript: "hello world", Confidence: 0.98}},
},
}
_ = conn.WriteJSON(final)
// wait briefly then close
time.Sleep(50 * time.Millisecond)
})
defer server.Close()
wsURL := "ws" + strings.TrimPrefix(server.URL, "http")
endpoint := &provider.EndpointConfig{BaseURL: wsURL, Path: ""}
adapter := NewDeepgramAdapter(endpoint, "test-api-key", "nova-3", "en")
ctx := context.Background()
if err := adapter.Start(ctx, ""); err != nil {
t.Fatalf("Start() error = %v", err)
}
// collect results
var results []TranscriptionResult
timeout := time.After(2 * time.Second)
loop:
for {
select {
case result, ok := <-adapter.Results():
if !ok {
break loop
}
results = append(results, result)
if result.IsFinal {
break loop
}
case <-timeout:
t.Fatal("timeout waiting for results")
}
}
adapter.Close()
wg.Wait()
// verify results
if len(results) < 2 {
t.Fatalf("expected at least 2 results, got %d", len(results))
}
// check interim
if results[0].Text != "hello" || results[0].IsFinal {
t.Errorf("interim result = %+v, want Text='hello', IsFinal=false", results[0])
}
// check final
found := false
for _, r := range results {
if r.Text == "hello world" && r.IsFinal {
found = true
break
}
}
if !found {
t.Errorf("did not find expected final result 'hello world'")
}
}
func TestDeepgramAdapter_SendsRawBinaryAudio(t *testing.T) {
receivedAudio := make(chan []byte, 1)
server := mockDeepgramServer(t, func(conn *websocket.Conn) {
// send metadata first
metadata := deepgramWSResponse{Type: "Metadata", Metadata: &deepgramMetadata{RequestID: "test-123"}}
_ = conn.WriteJSON(metadata)
// read audio chunk
msgType, data, err := conn.ReadMessage()
if err != nil {
return
}
if msgType != websocket.BinaryMessage {
t.Errorf("expected binary message, got %d", msgType)
}
receivedAudio <- data
// keep reading until close
for {
_, _, err := conn.ReadMessage()
if err != nil {
break
}
}
})
defer server.Close()
wsURL := "ws" + strings.TrimPrefix(server.URL, "http")
endpoint := &provider.EndpointConfig{BaseURL: wsURL, Path: ""}
adapter := NewDeepgramAdapter(endpoint, "test-api-key", "nova-3", "en")
ctx := context.Background()
if err := adapter.Start(ctx, ""); err != nil {
t.Fatalf("Start() error = %v", err)
}
// send audio chunk
testAudio := []byte{0x01, 0x02, 0x03, 0x04}
if err := adapter.SendChunk(testAudio); err != nil {
t.Errorf("SendChunk() error = %v", err)
}
// verify audio was received
select {
case audio := <-receivedAudio:
if string(audio) != string(testAudio) {
t.Errorf("received audio = %v, want %v", audio, testAudio)
}
case <-time.After(time.Second):
t.Error("timeout waiting for audio")
}
adapter.Close()
}
func TestDeepgramAdapter_HandlesError(t *testing.T) {
server := mockDeepgramServer(t, func(conn *websocket.Conn) {
// send error
errResp := deepgramWSResponse{
Type: "Error",
Error: &deepgramError{
Type: "AuthError",
Message: "Invalid API key",
},
}
data, _ := json.Marshal(errResp)
_ = conn.WriteMessage(websocket.TextMessage, data)
time.Sleep(50 * time.Millisecond)
})
defer server.Close()
wsURL := "ws" + strings.TrimPrefix(server.URL, "http")
endpoint := &provider.EndpointConfig{BaseURL: wsURL, Path: ""}
adapter := NewDeepgramAdapter(endpoint, "test-api-key", "nova-3", "en")
ctx := context.Background()
if err := adapter.Start(ctx, ""); err != nil {
t.Fatalf("Start() error = %v", err)
}
// wait for error result
select {
case result := <-adapter.Results():
if result.Error == nil {
t.Error("expected error result")
}
if !strings.Contains(result.Error.Error(), "Invalid API key") {
t.Errorf("error = %v, want to contain 'Invalid API key'", result.Error)
}
case <-time.After(time.Second):
t.Error("timeout waiting for error")
}
adapter.Close()
}
func TestDeepgramAdapter_ContextCancellation(t *testing.T) {
server := mockDeepgramServer(t, func(conn *websocket.Conn) {
// send metadata first so connection is established
metadata := deepgramWSResponse{Type: "Metadata", Metadata: &deepgramMetadata{RequestID: "test-123"}}
_ = conn.WriteJSON(metadata)
// just keep connection open until closed
for {
_, _, err := conn.ReadMessage()
if err != nil {
break
}
}
})
defer server.Close()
wsURL := "ws" + strings.TrimPrefix(server.URL, "http")
endpoint := &provider.EndpointConfig{BaseURL: wsURL, Path: ""}
adapter := NewDeepgramAdapter(endpoint, "test-api-key", "nova-3", "en")
ctx, cancel := context.WithCancel(context.Background())
if err := adapter.Start(ctx, ""); err != nil {
t.Fatalf("Start() error = %v", err)
}
// cancel context - this should trigger Close() to be called or at least stop SendChunk
cancel()
// SendChunk should return error after context cancelled
err := adapter.SendChunk([]byte("test"))
if err == nil {
// it's ok if first chunk after cancel succeeds - the context cancel is async
// but subsequent operations should fail
}
// Close should work even after context cancelled
if err := adapter.Close(); err != nil {
t.Errorf("Close() error = %v", err)
}
// results channel should be closed after Close()
select {
case _, ok := <-adapter.Results():
if ok {
// drain any remaining
for range adapter.Results() {
}
}
case <-time.After(2 * time.Second):
t.Error("timeout waiting for results channel to close")
}
}
+14 -1
View File
@@ -346,4 +346,17 @@ Started: Sun Feb 1 12:22:47 AM CET 2026
- DefaultModel(Transcription) returns 'nova-3' - DefaultModel(Transcription) returns 'nova-3'
- Registered in provider.init() - Registered in provider.init()
- Comprehensive test file created: deepgram_test.go - Comprehensive test file created: deepgram_test.go
- All tests passing, typecheck passes - All tests passing, typecheck passes
### Task 34: Create Deepgram StreamingAdapter
- Created `internal/transcriber/adapter_deepgram.go`
- DeepgramAdapter struct: endpoint, apiKey, model, language, conn, resultsCh, mutex, ctx/cancel, WaitGroup
- Start(): connects to wss://api.deepgram.com/v1/listen with Authorization: Token header
- Query params: model, language, encoding=linear16, sample_rate=16000, channels=1, interim_results=true, smart_format=true, punctuate=true
- Language conversion via language.ToProviderFormat(lang, "deepgram")
- SendChunk(): sends raw binary audio (websocket.BinaryMessage, not base64 like ElevenLabs)
- readLoop goroutine: parses Metadata, Results (interim + final), Error, UtteranceEnd, SpeechStarted messages
- Close(): cancels context, sends close frame, waits for reader goroutine
- Added reconnection logic (maxRetries=3, exponential backoff 1s, 2s, 4s) matching ElevenLabs pattern
- Comprehensive tests with mock WebSocket server
- All tests passing with -race flag, typecheck passes
+1 -1
View File
@@ -806,7 +806,7 @@
"Close() terminates cleanly", "Close() terminates cleanly",
"Typecheck passes" "Typecheck passes"
], ],
"passes": false "passes": true
}, },
{ {
"title": "Add reconnection logic to Deepgram StreamingAdapter", "title": "Add reconnection logic to Deepgram StreamingAdapter",