forked from intitni/CopilotForXcode
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathChatGPTService.swift
More file actions
345 lines (293 loc) · 11.4 KB
/
ChatGPTService.swift
File metadata and controls
345 lines (293 loc) · 11.4 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
import AsyncAlgorithms
import Foundation
import GPTEncoder
import Preferences
public protocol ChatGPTServiceType: ObservableObject {
var isReceivingMessage: Bool { get async }
var history: [ChatMessage] { get async }
func send(content: String, summary: String?) async throws -> AsyncThrowingStream<String, Error>
func stopReceivingMessage() async
func clearHistory() async
func mutateSystemPrompt(_ newPrompt: String) async
func mutateHistory(_ mutate: (inout [ChatMessage]) -> Void) async
func markReceivingMessage(_ receiving: Bool) async
}
public enum ChatGPTServiceError: Error, LocalizedError {
case endpointIncorrect
case responseInvalid
case otherError(String)
public var errorDescription: String? {
switch self {
case .endpointIncorrect:
return "ChatGPT endpoint is incorrect"
case .responseInvalid:
return "Response is invalid"
case let .otherError(content):
return content
}
}
}
public struct ChatGPTError: Error, Codable, LocalizedError {
public var error: ErrorContent
public init(error: ErrorContent) {
self.error = error
}
public struct ErrorContent: Codable {
public var message: String
public var type: String
public var param: String?
public var code: String?
public init(message: String, type: String, param: String? = nil, code: String? = nil) {
self.message = message
self.type = type
self.param = param
self.code = code
}
}
public var errorDescription: String? {
error.message
}
}
public actor ChatGPTService: ChatGPTServiceType {
public var systemPrompt: String
public var defaultTemperature: Double {
min(max(0, UserDefaults.shared.value(for: \.chatGPTTemperature)), 2)
}
var temperature: Double?
public var model: String {
let value = UserDefaults.shared.value(for: \.chatGPTModel)
if value.isEmpty { return "gpt-3.5-turbo" }
return value
}
var designatedProvider: ChatFeatureProvider?
public var endpoint: String {
switch designatedProvider ?? UserDefaults.shared.value(for: \.chatFeatureProvider) {
case .openAI:
let baseURL = UserDefaults.shared.value(for: \.openAIBaseURL)
if baseURL.isEmpty { return "https://api.openai.com/v1/chat/completions" }
return "\(baseURL)/v1/chat/completions"
case .azureOpenAI:
let baseURL = UserDefaults.shared.value(for: \.azureOpenAIBaseURL)
let deployment = UserDefaults.shared.value(for: \.azureChatGPTDeployment)
let version = "2023-05-15"
if baseURL.isEmpty { return "" }
return "\(baseURL)/openai/deployments/\(deployment)/chat/completions?api-version=\(version)"
}
}
public var apiKey: String {
switch designatedProvider ?? UserDefaults.shared.value(for: \.chatFeatureProvider) {
case .openAI:
return UserDefaults.shared.value(for: \.openAIAPIKey)
case .azureOpenAI:
return UserDefaults.shared.value(for: \.azureOpenAIAPIKey)
}
}
public var maxToken: Int {
UserDefaults.shared.value(for: \.chatGPTMaxToken)
}
public var history: [ChatMessage] = [] {
didSet { objectWillChange.send() }
}
public internal(set) var isReceivingMessage = false {
didSet { objectWillChange.send() }
}
var stop: [String]
var uuidGenerator: () -> String = { UUID().uuidString }
var cancelTask: Cancellable?
var buildCompletionStreamAPI: CompletionStreamAPIBuilder = OpenAICompletionStreamAPI.init
var buildCompletionAPI: CompletionAPIBuilder = OpenAICompletionAPI.init
public init(
systemPrompt: String = "",
temperature: Double? = nil,
stop: [String] = [],
designatedProvider: ChatFeatureProvider? = nil
) {
self.systemPrompt = systemPrompt
self.temperature = temperature
self.stop = stop
self.designatedProvider = designatedProvider
}
public func send(
content: String,
summary: String? = nil
) async throws -> AsyncThrowingStream<String, Error> {
guard !isReceivingMessage else { throw CancellationError() }
guard let url = URL(string: endpoint) else { throw ChatGPTServiceError.endpointIncorrect }
if !content.isEmpty || summary != nil {
let newMessage = ChatMessage(
id: uuidGenerator(),
role: .user,
content: content,
summary: summary
)
history.append(newMessage)
}
let (messages, remainingTokens) = combineHistoryWithSystemPrompt()
let requestBody = CompletionRequestBody(
model: model,
messages: messages,
temperature: temperature ?? defaultTemperature,
stream: true,
stop: stop.isEmpty ? nil : stop,
max_tokens: maxTokenForReply(model: model, remainingTokens: remainingTokens)
)
isReceivingMessage = true
let api = buildCompletionStreamAPI(
apiKey,
designatedProvider ?? UserDefaults.shared.value(for: \.chatFeatureProvider),
url,
requestBody
)
return AsyncThrowingStream<String, Error> { continuation in
Task {
do {
let (trunks, cancel) = try await api()
guard isReceivingMessage else {
continuation.finish()
return
}
cancelTask = cancel
for try await trunk in trunks {
guard let delta = trunk.choices.first?.delta else { continue }
if history.last?.id == trunk.id {
if let role = delta.role {
history[history.endIndex - 1].role = role
}
if let content = delta.content {
history[history.endIndex - 1].content.append(content)
}
} else {
history.append(.init(
id: trunk.id,
role: delta.role ?? .assistant,
content: delta.content ?? ""
))
}
if let content = delta.content {
continuation.yield(content)
}
try await Task.sleep(nanoseconds: 3_500_000)
}
continuation.finish()
isReceivingMessage = false
} catch let error as CancellationError {
isReceivingMessage = false
continuation.finish(throwing: error)
} catch let error as NSError where error.code == NSURLErrorCancelled {
isReceivingMessage = false
continuation.finish(throwing: error)
} catch {
history.append(.init(
role: .assistant,
content: error.localizedDescription
))
isReceivingMessage = false
continuation.finish(throwing: error)
}
}
}
}
public func sendAndWait(
content: String,
summary: String? = nil
) async throws -> String? {
guard !isReceivingMessage else { throw CancellationError() }
guard let url = URL(string: endpoint) else { throw ChatGPTServiceError.endpointIncorrect }
if !content.isEmpty || summary != nil {
let newMessage = ChatMessage(
id: uuidGenerator(),
role: .user,
content: content,
summary: summary
)
history.append(newMessage)
}
let (messages, remainingTokens) = combineHistoryWithSystemPrompt()
let requestBody = CompletionRequestBody(
model: model,
messages: messages,
temperature: temperature ?? defaultTemperature,
stream: true,
stop: stop.isEmpty ? nil : stop,
max_tokens: maxTokenForReply(model: model, remainingTokens: remainingTokens)
)
isReceivingMessage = true
defer { isReceivingMessage = false }
let api = buildCompletionAPI(
apiKey,
designatedProvider ?? UserDefaults.shared.value(for: \.chatFeatureProvider),
url,
requestBody
)
let response = try await api()
if let choice = response.choices.first {
history.append(.init(
id: response.id,
role: choice.message.role,
content: choice.message.content
))
return choice.message.content
}
return nil
}
public func stopReceivingMessage() {
cancelTask?()
cancelTask = nil
isReceivingMessage = false
}
public func clearHistory() {
stopReceivingMessage()
history = []
}
public func mutateSystemPrompt(_ newPrompt: String) {
systemPrompt = newPrompt
}
public func mutateHistory(_ mutate: (inout [ChatMessage]) -> Void) async {
mutate(&history)
}
public func markReceivingMessage(_ receiving: Bool) {
isReceivingMessage = receiving
}
}
extension ChatGPTService {
func changeBuildCompletionStreamAPI(_ builder: @escaping CompletionStreamAPIBuilder) {
buildCompletionStreamAPI = builder
}
func changeUUIDGenerator(_ generator: @escaping () -> String) {
uuidGenerator = generator
}
func combineHistoryWithSystemPrompt(
minimumReplyTokens: Int = 300,
maxNumberOfMessages: Int = UserDefaults.shared.value(for: \.chatGPTMaxMessageCount),
maxTokens: Int = UserDefaults.shared.value(for: \.chatGPTMaxToken),
encoder: TokenEncoder = GPTEncoder()
)
-> (messages: [CompletionRequestBody.Message], remainingTokens: Int)
{
var all: [CompletionRequestBody.Message] = []
var allTokensCount = encoder.encode(text: systemPrompt).count
for (index, message) in history.enumerated().reversed() {
if maxNumberOfMessages > 0, all.count >= maxNumberOfMessages { break }
if message.content.isEmpty { continue }
let tokensCount = message.tokensCount ?? encoder.encode(text: message.content).count
history[index].tokensCount = tokensCount
if tokensCount + allTokensCount > maxTokens - minimumReplyTokens {
break
}
allTokensCount += tokensCount
all.append(.init(role: message.role, content: message.content))
}
if !systemPrompt.isEmpty {
all.append(.init(role: .system, content: systemPrompt))
}
return (all.reversed(), max(minimumReplyTokens, maxTokens - allTokensCount))
}
}
protocol TokenEncoder {
func encode(text: String) -> [Int]
}
extension GPTEncoder: TokenEncoder {}
func maxTokenForReply(model: String, remainingTokens: Int) -> Int {
guard let model = ChatGPTModel(rawValue: model) else { return remainingTokens }
return min(model.maxToken / 2, remainingTokens)
}