44 "context"
55 "encoding/json"
66 "errors"
7- "fmt"
87 "net/http"
98 "net/url"
109 "strconv"
@@ -15,10 +14,12 @@ import (
1514)
1615
1716var (
18- // ErrSessionClosed is returned when attempting to write to a closed session.
17+ // ErrSessionClosed is returned when attempting to write to a closed
18+ // session.
1919 ErrSessionClosed = errors .New ("session closed" )
2020
21- // ErrDisconnected is returned when attempting to write to a disconnected client.
21+ // ErrDisconnected is returned when attempting to write to a disconnected
22+ // client.
2223 ErrDisconnected = errors .New ("client is disconnected" )
2324)
2425
@@ -72,8 +73,10 @@ type RealTimeBaseTranscript struct {
7273 // The partial transcript for your audio
7374 Text string `json:"text"`
7475
75- // An array of objects, with the information for each word in the transcription text.
76- // Includes the start and end time of the word in milliseconds, the confidence score of the word, and the text, which is the word itself.
76+ // An array of objects, with the information for each word in the
77+ // transcription text. Includes the start and end time of the word in
78+ // milliseconds, the confidence score of the word, and the text, which is
79+ // the word itself.
7780 Words []Word `json:"words"`
7881}
7982
@@ -116,8 +119,10 @@ var DefaultSampleRate = 16_000
116119type RealTimeClient struct {
117120 baseURL * url.URL
118121 apiKey string
122+ token string
119123
120- conn * websocket.Conn
124+ conn * websocket.Conn
125+ httpClient * http.Client
121126
122127 mtx sync.RWMutex
123128 sessionOpen bool
@@ -126,6 +131,10 @@ type RealTimeClient struct {
126131 done chan bool
127132
128133 handler RealTimeHandler
134+
135+ sampleRate int
136+ encoding RealTimeEncoding
137+ wordBoost []string
129138}
130139
131140func (c * RealTimeClient ) isSessionOpen () bool {
@@ -148,7 +157,8 @@ type RealTimeError struct {
148157
149158type RealTimeClientOption func (* RealTimeClient )
150159
151- // WithRealTimeBaseURL sets the API endpoint used by the client. Mainly used for testing.
160+ // WithRealTimeBaseURL sets the API endpoint used by the client. Mainly used for
161+ // testing.
152162func WithRealTimeBaseURL (rawurl string ) RealTimeClientOption {
153163 return func (c * RealTimeClient ) {
154164 if u , err := url .Parse (rawurl ); err == nil {
@@ -157,12 +167,22 @@ func WithRealTimeBaseURL(rawurl string) RealTimeClientOption {
157167 }
158168}
159169
170+ // WithRealTimeAuthToken configures the client to authenticate using an
171+ // AssemblyAI API key.
160172func WithRealTimeAPIKey (apiKey string ) RealTimeClientOption {
161173 return func (rtc * RealTimeClient ) {
162174 rtc .apiKey = apiKey
163175 }
164176}
165177
178+ // WithRealTimeAuthToken configures the client to authenticate using a temporary
179+ // token generated using [CreateTemporaryToken].
180+ func WithRealTimeAuthToken (token string ) RealTimeClientOption {
181+ return func (rtc * RealTimeClient ) {
182+ rtc .token = token
183+ }
184+ }
185+
166186func WithHandler (handler RealTimeHandler ) RealTimeClientOption {
167187 return func (rtc * RealTimeClient ) {
168188 rtc .handler = handler
@@ -171,24 +191,13 @@ func WithHandler(handler RealTimeHandler) RealTimeClientOption {
171191
172192func WithRealTimeSampleRate (sampleRate int ) RealTimeClientOption {
173193 return func (rtc * RealTimeClient ) {
174- if sampleRate > 0 {
175- vs := rtc .baseURL .Query ()
176- vs .Set ("sample_rate" , strconv .Itoa (sampleRate ))
177- rtc .baseURL .RawQuery = vs .Encode ()
178- }
194+ rtc .sampleRate = sampleRate
179195 }
180196}
181197
182198func WithRealTimeWordBoost (wordBoost []string ) RealTimeClientOption {
183199 return func (rtc * RealTimeClient ) {
184- vs := rtc .baseURL .Query ()
185-
186- if len (wordBoost ) > 0 {
187- b , _ := json .Marshal (wordBoost )
188- vs .Set ("word_boost" , string (b ))
189- }
190-
191- rtc .baseURL .RawQuery = vs .Encode ()
200+ rtc .wordBoost = wordBoost
192201 }
193202}
194203
@@ -205,26 +214,26 @@ const (
205214
206215func WithRealTimeEncoding (encoding RealTimeEncoding ) RealTimeClientOption {
207216 return func (rtc * RealTimeClient ) {
208- vs := rtc .baseURL .Query ()
209- vs .Set ("encoding" , string (encoding ))
210- rtc .baseURL .RawQuery = vs .Encode ()
217+ rtc .encoding = encoding
211218 }
212219}
213220
214221func NewRealTimeClientWithOptions (options ... RealTimeClientOption ) * RealTimeClient {
215222 client := & RealTimeClient {
216223 baseURL : & url.URL {
217- Scheme : "wss" ,
218- Host : "api.assemblyai.com" ,
219- Path : "/v2/realtime/ws" ,
220- RawQuery : fmt .Sprintf ("sample_rate=%v" , DefaultSampleRate ),
224+ Scheme : "wss" ,
225+ Host : "api.assemblyai.com" ,
226+ Path : "/v2/realtime/ws" ,
221227 },
228+ httpClient : & http.Client {},
222229 }
223230
224231 for _ , option := range options {
225232 option (client )
226233 }
227234
235+ client .baseURL .RawQuery = client .queryFromOptions ()
236+
228237 return client
229238}
230239
@@ -261,7 +270,6 @@ func NewRealTimeClient(apiKey string, handler RealTimeHandler) *RealTimeClient {
261270// Closes the any open WebSocket connection in case of errors.
262271func (c * RealTimeClient ) Connect (ctx context.Context ) error {
263272 header := make (http.Header )
264- header .Set ("Authorization" , c .apiKey )
265273
266274 opts := & websocket.DialOptions {
267275 HTTPHeader : header ,
@@ -360,6 +368,33 @@ func (c *RealTimeClient) Connect(ctx context.Context) error {
360368 return nil
361369}
362370
371+ func (c * RealTimeClient ) queryFromOptions () string {
372+ values := url.Values {}
373+
374+ // Temporary token
375+ if c .token != "" {
376+ values .Set ("token" , c .token )
377+ }
378+
379+ // Sample rate
380+ if c .sampleRate > 0 {
381+ values .Set ("sample_rate" , strconv .Itoa (c .sampleRate ))
382+ }
383+
384+ // Encoding
385+ if c .encoding != "" {
386+ values .Set ("encoding" , string (c .encoding ))
387+ }
388+
389+ // Word boost
390+ if len (c .wordBoost ) > 0 {
391+ b , _ := json .Marshal (c .wordBoost )
392+ values .Set ("word_boost" , string (b ))
393+ }
394+
395+ return values .Encode ()
396+ }
397+
363398// Disconnect sends the terminate_session message and waits for the server to
364399// send a SessionTerminated message before closing the connection.
365400func (c * RealTimeClient ) Disconnect (ctx context.Context , waitForSessionTermination bool ) error {
@@ -405,3 +440,30 @@ func (c *RealTimeClient) SetEndUtteranceSilenceThreshold(ctx context.Context, th
405440 EndUtteranceSilenceThreshold : threshold ,
406441 })
407442}
443+
444+ // RealTimeService groups operations related to the real-time transcription.
445+ type RealTimeService struct {
446+ client * Client
447+ }
448+
449+ // CreateTemporaryToken creates a temporary token that can be used to
450+ // authenticate a real-time client.
451+ func (svc * RealTimeService ) CreateTemporaryToken (ctx context.Context , expiresIn int64 ) (* RealtimeTemporaryTokenResponse , error ) {
452+ params := & CreateRealtimeTemporaryTokenParams {
453+ ExpiresIn : Int64 (expiresIn ),
454+ }
455+
456+ req , err := svc .client .newJSONRequest ("POST" , "/v2/realtime/token" , params )
457+ if err != nil {
458+ return nil , err
459+ }
460+
461+ var tokenResponse RealtimeTemporaryTokenResponse
462+ resp , err := svc .client .do (ctx , req , & tokenResponse )
463+ if err != nil {
464+ return nil , err
465+ }
466+ defer resp .Body .Close ()
467+
468+ return & tokenResponse , nil
469+ }
0 commit comments