1
This commit is contained in:
413
keyBoard/Class/AiTalk/VM/DeepgramWebSocketClient.m
Normal file
413
keyBoard/Class/AiTalk/VM/DeepgramWebSocketClient.m
Normal file
@@ -0,0 +1,413 @@
|
||||
//
|
||||
// DeepgramWebSocketClient.m
|
||||
// keyBoard
|
||||
//
|
||||
// Created by Mac on 2026/1/21.
|
||||
//
|
||||
|
||||
#import "DeepgramWebSocketClient.h"
|
||||
|
||||
static NSString *const kDeepgramWebSocketClientErrorDomain =
|
||||
@"DeepgramWebSocketClient";
|
||||
|
||||
@interface DeepgramWebSocketClient () <NSURLSessionWebSocketDelegate>
|
||||
|
||||
@property(nonatomic, strong) NSURLSession *urlSession;
|
||||
@property(nonatomic, strong) NSURLSessionWebSocketTask *webSocketTask;
|
||||
@property(nonatomic, strong) dispatch_queue_t networkQueue;
|
||||
@property(nonatomic, assign) BOOL connected;
|
||||
@property(nonatomic, assign) BOOL audioSendingEnabled;
|
||||
|
||||
@end
|
||||
|
||||
@implementation DeepgramWebSocketClient
|
||||
|
||||
- (instancetype)init {
|
||||
self = [super init];
|
||||
if (self) {
|
||||
_networkQueue = dispatch_queue_create("com.keyboard.aitalk.deepgram.ws",
|
||||
DISPATCH_QUEUE_SERIAL);
|
||||
_serverURL = @"wss://api.deepgram.com/v1/listen";
|
||||
_encoding = @"linear16";
|
||||
_sampleRate = 16000.0;
|
||||
_channels = 1;
|
||||
_punctuate = YES;
|
||||
_smartFormat = YES;
|
||||
_interimResults = YES;
|
||||
_audioSendingEnabled = NO;
|
||||
}
|
||||
return self;
|
||||
}
|
||||
|
||||
- (void)dealloc {
|
||||
[self disconnect];
|
||||
}
|
||||
|
||||
#pragma mark - Public Methods
|
||||
|
||||
- (void)connect {
|
||||
dispatch_async(self.networkQueue, ^{
|
||||
[self disconnectInternal];
|
||||
|
||||
if (self.apiKey.length == 0) {
|
||||
[self reportErrorWithMessage:@"Deepgram API key is required"];
|
||||
return;
|
||||
}
|
||||
|
||||
NSURL *url = [self buildURL];
|
||||
if (!url) {
|
||||
[self reportErrorWithMessage:@"Invalid Deepgram URL"];
|
||||
return;
|
||||
}
|
||||
|
||||
NSLog(@"[DeepgramWebSocketClient] Connecting: %@", url.absoluteString);
|
||||
|
||||
NSURLSessionConfiguration *config =
|
||||
[NSURLSessionConfiguration defaultSessionConfiguration];
|
||||
config.timeoutIntervalForRequest = 30;
|
||||
config.timeoutIntervalForResource = 300;
|
||||
|
||||
self.urlSession = [NSURLSession sessionWithConfiguration:config
|
||||
delegate:self
|
||||
delegateQueue:nil];
|
||||
|
||||
NSMutableURLRequest *request = [NSMutableURLRequest requestWithURL:url];
|
||||
[request setValue:[NSString stringWithFormat:@"Token %@", self.apiKey]
|
||||
forHTTPHeaderField:@"Authorization"];
|
||||
|
||||
self.webSocketTask = [self.urlSession webSocketTaskWithRequest:request];
|
||||
[self.webSocketTask resume];
|
||||
[self receiveMessage];
|
||||
});
|
||||
}
|
||||
|
||||
- (void)disconnect {
|
||||
dispatch_async(self.networkQueue, ^{
|
||||
BOOL shouldNotify = self.webSocketTask != nil;
|
||||
if (shouldNotify) {
|
||||
NSLog(@"[DeepgramWebSocketClient] Disconnect requested");
|
||||
}
|
||||
[self disconnectInternal];
|
||||
if (shouldNotify) {
|
||||
[self notifyDisconnect:nil];
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
- (void)sendAudioPCMFrame:(NSData *)pcmFrame {
|
||||
if (!self.connected || !self.webSocketTask || pcmFrame.length == 0) {
|
||||
return;
|
||||
}
|
||||
|
||||
dispatch_async(self.networkQueue, ^{
|
||||
if (!self.audioSendingEnabled) {
|
||||
return;
|
||||
}
|
||||
if (!self.connected || !self.webSocketTask) {
|
||||
return;
|
||||
}
|
||||
|
||||
NSURLSessionWebSocketMessage *message =
|
||||
[[NSURLSessionWebSocketMessage alloc] initWithData:pcmFrame];
|
||||
[self.webSocketTask
|
||||
sendMessage:message
|
||||
completionHandler:^(NSError *_Nullable error) {
|
||||
if (error) {
|
||||
[self reportError:error];
|
||||
} else {
|
||||
NSLog(@"[DeepgramWebSocketClient] Sent audio frame: %lu bytes",
|
||||
(unsigned long)pcmFrame.length);
|
||||
}
|
||||
}];
|
||||
});
|
||||
}
|
||||
|
||||
- (void)finish {
|
||||
NSLog(@"[DeepgramWebSocketClient] Sending CloseStream");
|
||||
[self sendJSON:@{ @"type" : @"CloseStream" }];
|
||||
}
|
||||
|
||||
- (void)sendKeepAlive {
|
||||
if (!self.connected || !self.webSocketTask) {
|
||||
return;
|
||||
}
|
||||
[self sendJSON:@{ @"type" : @"KeepAlive" }];
|
||||
}
|
||||
|
||||
- (void)enableAudioSending {
|
||||
dispatch_async(self.networkQueue, ^{
|
||||
self.audioSendingEnabled = YES;
|
||||
});
|
||||
}
|
||||
|
||||
- (void)disableAudioSending {
|
||||
dispatch_async(self.networkQueue, ^{
|
||||
self.audioSendingEnabled = NO;
|
||||
});
|
||||
}
|
||||
|
||||
#pragma mark - Private Methods
|
||||
|
||||
- (NSURL *)buildURL {
|
||||
if (self.serverURL.length == 0) {
|
||||
return nil;
|
||||
}
|
||||
|
||||
NSURLComponents *components =
|
||||
[NSURLComponents componentsWithString:self.serverURL];
|
||||
if (!components) {
|
||||
return nil;
|
||||
}
|
||||
|
||||
NSMutableArray<NSURLQueryItem *> *items =
|
||||
components.queryItems.mutableCopy ?: [NSMutableArray array];
|
||||
|
||||
[self upsertQueryItemWithName:@"model" value:self.model items:items];
|
||||
[self upsertQueryItemWithName:@"language" value:self.language items:items];
|
||||
|
||||
[self upsertQueryItemWithName:@"punctuate"
|
||||
value:(self.punctuate ? @"true" : @"false")
|
||||
items:items];
|
||||
[self upsertQueryItemWithName:@"smart_format"
|
||||
value:(self.smartFormat ? @"true" : @"false")
|
||||
items:items];
|
||||
[self upsertQueryItemWithName:@"interim_results"
|
||||
value:(self.interimResults ? @"true" : @"false")
|
||||
items:items];
|
||||
|
||||
[self upsertQueryItemWithName:@"encoding" value:self.encoding items:items];
|
||||
[self upsertQueryItemWithName:@"sample_rate"
|
||||
value:[NSString stringWithFormat:@"%.0f",
|
||||
self.sampleRate]
|
||||
items:items];
|
||||
[self upsertQueryItemWithName:@"channels"
|
||||
value:[NSString stringWithFormat:@"%d", self.channels]
|
||||
items:items];
|
||||
|
||||
components.queryItems = items;
|
||||
return components.URL;
|
||||
}
|
||||
|
||||
- (void)upsertQueryItemWithName:(NSString *)name
|
||||
value:(NSString *)value
|
||||
items:(NSMutableArray<NSURLQueryItem *> *)items {
|
||||
if (name.length == 0 || value.length == 0) {
|
||||
return;
|
||||
}
|
||||
|
||||
for (NSUInteger i = 0; i < items.count; i++) {
|
||||
NSURLQueryItem *item = items[i];
|
||||
if ([item.name isEqualToString:name]) {
|
||||
items[i] = [NSURLQueryItem queryItemWithName:name value:value];
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
||||
[items addObject:[NSURLQueryItem queryItemWithName:name value:value]];
|
||||
}
|
||||
|
||||
- (void)sendJSON:(NSDictionary *)dict {
|
||||
if (!self.webSocketTask) {
|
||||
return;
|
||||
}
|
||||
|
||||
NSError *jsonError = nil;
|
||||
NSData *jsonData = [NSJSONSerialization dataWithJSONObject:dict
|
||||
options:0
|
||||
error:&jsonError];
|
||||
if (jsonError) {
|
||||
[self reportError:jsonError];
|
||||
return;
|
||||
}
|
||||
|
||||
NSString *jsonString =
|
||||
[[NSString alloc] initWithData:jsonData
|
||||
encoding:NSUTF8StringEncoding];
|
||||
if (!jsonString) {
|
||||
[self reportErrorWithMessage:@"Failed to encode JSON message"];
|
||||
return;
|
||||
}
|
||||
|
||||
dispatch_async(self.networkQueue, ^{
|
||||
NSURLSessionWebSocketMessage *message =
|
||||
[[NSURLSessionWebSocketMessage alloc] initWithString:jsonString];
|
||||
[self.webSocketTask
|
||||
sendMessage:message
|
||||
completionHandler:^(NSError *_Nullable error) {
|
||||
if (error) {
|
||||
[self reportError:error];
|
||||
}
|
||||
}];
|
||||
});
|
||||
}
|
||||
|
||||
- (void)receiveMessage {
|
||||
if (!self.webSocketTask) {
|
||||
return;
|
||||
}
|
||||
|
||||
__weak typeof(self) weakSelf = self;
|
||||
[self.webSocketTask receiveMessageWithCompletionHandler:^(
|
||||
NSURLSessionWebSocketMessage *_Nullable message,
|
||||
NSError *_Nullable error) {
|
||||
__strong typeof(weakSelf) strongSelf = weakSelf;
|
||||
if (!strongSelf) {
|
||||
return;
|
||||
}
|
||||
|
||||
if (error) {
|
||||
if (error.code != NSURLErrorCancelled && error.code != 57) {
|
||||
[strongSelf notifyDisconnect:error];
|
||||
[strongSelf disconnectInternal];
|
||||
}
|
||||
return;
|
||||
}
|
||||
|
||||
if (message.type == NSURLSessionWebSocketMessageTypeString) {
|
||||
NSLog(@"[DeepgramWebSocketClient] Received text: %@", message.string);
|
||||
[strongSelf handleTextMessage:message.string];
|
||||
} else if (message.type == NSURLSessionWebSocketMessageTypeData) {
|
||||
NSLog(@"[DeepgramWebSocketClient] Received binary: %lu bytes",
|
||||
(unsigned long)message.data.length);
|
||||
[strongSelf handleBinaryMessage:message.data];
|
||||
}
|
||||
|
||||
[strongSelf receiveMessage];
|
||||
}];
|
||||
}
|
||||
|
||||
- (void)handleTextMessage:(NSString *)text {
|
||||
if (text.length == 0) {
|
||||
return;
|
||||
}
|
||||
|
||||
NSData *data = [text dataUsingEncoding:NSUTF8StringEncoding];
|
||||
if (!data) {
|
||||
return;
|
||||
}
|
||||
|
||||
NSError *jsonError = nil;
|
||||
NSDictionary *json = [NSJSONSerialization JSONObjectWithData:data
|
||||
options:0
|
||||
error:&jsonError];
|
||||
if (jsonError) {
|
||||
[self reportError:jsonError];
|
||||
return;
|
||||
}
|
||||
|
||||
NSString *errorMessage = json[@"error"];
|
||||
if (errorMessage.length > 0) {
|
||||
[self reportErrorWithMessage:errorMessage];
|
||||
return;
|
||||
}
|
||||
|
||||
NSDictionary *channel = json[@"channel"];
|
||||
if (![channel isKindOfClass:[NSDictionary class]]) {
|
||||
return;
|
||||
}
|
||||
|
||||
NSArray *alternatives = channel[@"alternatives"];
|
||||
if (![alternatives isKindOfClass:[NSArray class]] || alternatives.count == 0) {
|
||||
return;
|
||||
}
|
||||
|
||||
NSDictionary *firstAlt = alternatives.firstObject;
|
||||
NSString *transcript = firstAlt[@"transcript"] ?: @"";
|
||||
BOOL isFinal = [json[@"is_final"] boolValue] ||
|
||||
[json[@"speech_final"] boolValue];
|
||||
|
||||
if (transcript.length == 0) {
|
||||
return;
|
||||
}
|
||||
|
||||
dispatch_async(dispatch_get_main_queue(), ^{
|
||||
if (isFinal) {
|
||||
if ([self.delegate respondsToSelector:@selector
|
||||
(deepgramClientDidReceiveFinalTranscript:)]) {
|
||||
[self.delegate deepgramClientDidReceiveFinalTranscript:transcript];
|
||||
}
|
||||
} else {
|
||||
if ([self.delegate respondsToSelector:@selector
|
||||
(deepgramClientDidReceiveInterimTranscript:)]) {
|
||||
[self.delegate deepgramClientDidReceiveInterimTranscript:transcript];
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
- (void)handleBinaryMessage:(NSData *)data {
|
||||
}
|
||||
|
||||
- (void)disconnectInternal {
|
||||
self.connected = NO;
|
||||
self.audioSendingEnabled = NO;
|
||||
|
||||
if (self.webSocketTask) {
|
||||
[self.webSocketTask
|
||||
cancelWithCloseCode:NSURLSessionWebSocketCloseCodeNormalClosure
|
||||
reason:nil];
|
||||
self.webSocketTask = nil;
|
||||
}
|
||||
|
||||
if (self.urlSession) {
|
||||
[self.urlSession invalidateAndCancel];
|
||||
self.urlSession = nil;
|
||||
}
|
||||
}
|
||||
|
||||
- (void)reportError:(NSError *)error {
|
||||
dispatch_async(dispatch_get_main_queue(), ^{
|
||||
if ([self.delegate respondsToSelector:@selector(deepgramClientDidFail:)]) {
|
||||
[self.delegate deepgramClientDidFail:error];
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
- (void)reportErrorWithMessage:(NSString *)message {
|
||||
NSError *error = [NSError errorWithDomain:kDeepgramWebSocketClientErrorDomain
|
||||
code:-1
|
||||
userInfo:@{
|
||||
NSLocalizedDescriptionKey : message ?: @""
|
||||
}];
|
||||
[self reportError:error];
|
||||
}
|
||||
|
||||
- (void)notifyDisconnect:(NSError *_Nullable)error {
|
||||
self.connected = NO;
|
||||
|
||||
dispatch_async(dispatch_get_main_queue(), ^{
|
||||
if ([self.delegate respondsToSelector:@selector
|
||||
(deepgramClientDidDisconnect:)]) {
|
||||
[self.delegate deepgramClientDidDisconnect:error];
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
#pragma mark - NSURLSessionWebSocketDelegate
|
||||
|
||||
- (void)URLSession:(NSURLSession *)session
|
||||
webSocketTask:(NSURLSessionWebSocketTask *)webSocketTask
|
||||
didOpenWithProtocol:(NSString *)protocol {
|
||||
self.connected = YES;
|
||||
NSLog(@"[DeepgramWebSocketClient] Connected");
|
||||
dispatch_async(dispatch_get_main_queue(), ^{
|
||||
if ([self.delegate respondsToSelector:@selector(deepgramClientDidConnect)]) {
|
||||
[self.delegate deepgramClientDidConnect];
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
- (void)URLSession:(NSURLSession *)session
|
||||
webSocketTask:(NSURLSessionWebSocketTask *)webSocketTask
|
||||
didCloseWithCode:(NSURLSessionWebSocketCloseCode)closeCode
|
||||
reason:(NSData *)reason {
|
||||
if (!self.webSocketTask) {
|
||||
return;
|
||||
}
|
||||
NSLog(@"[DeepgramWebSocketClient] Closed with code: %ld",
|
||||
(long)closeCode);
|
||||
[self notifyDisconnect:nil];
|
||||
[self disconnectInternal];
|
||||
}
|
||||
|
||||
@end
|
||||
Reference in New Issue
Block a user