Files
keyboard/keyBoard/Class/AiTalk/VM/ASRStreamClient.m

272 lines
8.1 KiB
Mathematica
Raw Normal View History

//
// ASRStreamClient.m
// keyBoard
//
// Created by Mac on 2026/1/15.
//
#import "ASRStreamClient.h"
#import "AudioCaptureManager.h"
@interface ASRStreamClient () <NSURLSessionWebSocketDelegate>
@property(nonatomic, strong) NSURLSession *urlSession;
@property(nonatomic, strong) NSURLSessionWebSocketTask *webSocketTask;
@property(nonatomic, copy) NSString *currentSessionId;
@property(nonatomic, strong) dispatch_queue_t networkQueue;
@property(nonatomic, assign) BOOL connected;
@end
@implementation ASRStreamClient
- (instancetype)init {
self = [super init];
if (self) {
_networkQueue = dispatch_queue_create("com.keyboard.aitalk.asr.network",
DISPATCH_QUEUE_SERIAL);
// TODO: ASR
_serverURL = @"wss://your-asr-server.com/ws/asr";
}
return self;
}
- (void)dealloc {
2026-01-30 21:38:58 +08:00
[self cancelInternal];
}
#pragma mark - Public Methods
- (void)startWithSessionId:(NSString *)sessionId {
dispatch_async(self.networkQueue, ^{
[self cancelInternal];
self.currentSessionId = sessionId;
// WebSocket
NSURL *url = [NSURL URLWithString:self.serverURL];
NSURLSessionConfiguration *config =
[NSURLSessionConfiguration defaultSessionConfiguration];
config.timeoutIntervalForRequest = 30;
config.timeoutIntervalForResource = 300;
self.urlSession = [NSURLSession sessionWithConfiguration:config
delegate:self
delegateQueue:nil];
self.webSocketTask = [self.urlSession webSocketTaskWithURL:url];
[self.webSocketTask resume];
// start
NSDictionary *startMessage = @{
@"type" : @"start",
@"sessionId" : sessionId,
@"format" : @"pcm_s16le",
@"sampleRate" : @(kAudioSampleRate),
@"channels" : @(kAudioChannels)
};
NSError *jsonError = nil;
NSData *jsonData = [NSJSONSerialization dataWithJSONObject:startMessage
options:0
error:&jsonError];
if (jsonError) {
[self reportError:jsonError];
return;
}
NSString *jsonString = [[NSString alloc] initWithData:jsonData
encoding:NSUTF8StringEncoding];
NSURLSessionWebSocketMessage *message =
[[NSURLSessionWebSocketMessage alloc] initWithString:jsonString];
[self.webSocketTask
sendMessage:message
completionHandler:^(NSError *_Nullable error) {
if (error) {
[self reportError:error];
} else {
self.connected = YES;
[self receiveMessage];
NSLog(@"[ASRStreamClient] Started session: %@", sessionId);
}
}];
});
}
- (void)sendAudioPCMFrame:(NSData *)pcmFrame {
if (!self.connected || !self.webSocketTask) {
return;
}
dispatch_async(self.networkQueue, ^{
NSURLSessionWebSocketMessage *message =
[[NSURLSessionWebSocketMessage alloc] initWithData:pcmFrame];
[self.webSocketTask sendMessage:message
completionHandler:^(NSError *_Nullable error) {
if (error) {
NSLog(@"[ASRStreamClient] Failed to send audio frame: %@",
error.localizedDescription);
}
}];
});
}
- (void)finalize {
if (!self.connected || !self.webSocketTask) {
return;
}
dispatch_async(self.networkQueue, ^{
NSDictionary *finalizeMessage =
@{@"type" : @"finalize", @"sessionId" : self.currentSessionId ?: @""};
NSError *jsonError = nil;
NSData *jsonData = [NSJSONSerialization dataWithJSONObject:finalizeMessage
options:0
error:&jsonError];
if (jsonError) {
[self reportError:jsonError];
return;
}
NSString *jsonString = [[NSString alloc] initWithData:jsonData
encoding:NSUTF8StringEncoding];
NSURLSessionWebSocketMessage *message =
[[NSURLSessionWebSocketMessage alloc] initWithString:jsonString];
[self.webSocketTask sendMessage:message
completionHandler:^(NSError *_Nullable error) {
if (error) {
[self reportError:error];
} else {
NSLog(@"[ASRStreamClient] Sent finalize for session: %@",
self.currentSessionId);
}
}];
});
}
- (void)cancel {
dispatch_async(self.networkQueue, ^{
[self cancelInternal];
});
}
#pragma mark - Private Methods
- (void)cancelInternal {
self.connected = NO;
if (self.webSocketTask) {
[self.webSocketTask cancel];
self.webSocketTask = nil;
}
if (self.urlSession) {
[self.urlSession invalidateAndCancel];
self.urlSession = nil;
}
self.currentSessionId = nil;
}
- (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 != 57 && error.code != NSURLErrorCancelled) {
[strongSelf reportError:error];
}
return;
}
if (message.type == NSURLSessionWebSocketMessageTypeString) {
[strongSelf handleTextMessage:message.string];
}
//
[strongSelf receiveMessage];
}];
}
- (void)handleTextMessage:(NSString *)text {
NSData *data = [text dataUsingEncoding:NSUTF8StringEncoding];
NSError *jsonError = nil;
NSDictionary *json = [NSJSONSerialization JSONObjectWithData:data
options:0
error:&jsonError];
if (jsonError) {
NSLog(@"[ASRStreamClient] Failed to parse message: %@", text);
return;
}
NSString *type = json[@"type"];
if ([type isEqualToString:@"partial"]) {
NSString *partialText = json[@"text"] ?: @"";
dispatch_async(dispatch_get_main_queue(), ^{
if ([self.delegate
respondsToSelector:@selector(asrClientDidReceivePartialText:)]) {
[self.delegate asrClientDidReceivePartialText:partialText];
}
});
} else if ([type isEqualToString:@"final"]) {
NSString *finalText = json[@"text"] ?: @"";
dispatch_async(dispatch_get_main_queue(), ^{
if ([self.delegate
respondsToSelector:@selector(asrClientDidReceiveFinalText:)]) {
[self.delegate asrClientDidReceiveFinalText:finalText];
}
});
//
[self cancelInternal];
} else if ([type isEqualToString:@"error"]) {
NSInteger code = [json[@"code"] integerValue];
NSString *message = json[@"message"] ?: @"Unknown error";
NSError *error =
[NSError errorWithDomain:@"ASRStreamClient"
code:code
userInfo:@{NSLocalizedDescriptionKey : message}];
[self reportError:error];
}
}
- (void)reportError:(NSError *)error {
dispatch_async(dispatch_get_main_queue(), ^{
if ([self.delegate respondsToSelector:@selector(asrClientDidFail:)]) {
[self.delegate asrClientDidFail:error];
}
});
}
#pragma mark - NSURLSessionWebSocketDelegate
- (void)URLSession:(NSURLSession *)session
webSocketTask:(NSURLSessionWebSocketTask *)webSocketTask
didOpenWithProtocol:(NSString *)protocol {
NSLog(@"[ASRStreamClient] WebSocket connected with protocol: %@", protocol);
}
- (void)URLSession:(NSURLSession *)session
webSocketTask:(NSURLSessionWebSocketTask *)webSocketTask
didCloseWithCode:(NSURLSessionWebSocketCloseCode)closeCode
reason:(NSData *)reason {
NSLog(@"[ASRStreamClient] WebSocket closed with code: %ld", (long)closeCode);
self.connected = NO;
}
@end