78d3235ab6
- Updated HTML assets for improved loading. - Integrated token usage tracking in chat processing, appending usage details to transcripts. - Enhanced OpenAIProvider to include usage data in chat completion responses. - Implemented asynchronous retrieval of recent token usage history. - Added utility functions for managing transcript files and parsing usage data. - Updated UI components to reflect changes in usage status handling. - Ensured consistent usage status definitions across the application.
69 lines
2.3 KiB
TypeScript
69 lines
2.3 KiB
TypeScript
import { BaseProvider, ChatOptions, GatewayChatMessage } from "./BaseProvider";
|
|
|
|
import OpenAI from "openai";
|
|
import logManager from "@electron/service/logger"
|
|
|
|
function _transformChunk(chunk: OpenAI.Chat.Completions.ChatCompletionChunk): UniversalChunk {
|
|
const choice = chunk.choices[0];
|
|
const usage = (chunk as any).usage;
|
|
return {
|
|
isEnd: choice?.finish_reason != null || (chunk.choices.length === 0 && usage != null),
|
|
result: choice?.delta?.content ?? '',
|
|
usage: usage ?? undefined,
|
|
}
|
|
}
|
|
|
|
export class OpenAIProvider extends BaseProvider {
|
|
private client: OpenAI;
|
|
|
|
constructor(apiKey: string, baseURL: string, headers?: Record<string, string>) {
|
|
super();
|
|
this.client = new OpenAI({ apiKey, baseURL, defaultHeaders: headers });
|
|
}
|
|
|
|
async chat(messages: GatewayChatMessage[], model: string, options?: ChatOptions): Promise<AsyncIterable<UniversalChunk>> {
|
|
const startTime = Date.now();
|
|
|
|
const lastMessage = messages[messages.length - 1];
|
|
|
|
logManager.logApiRequest('chat.completions.create', {
|
|
model,
|
|
lastMessage: lastMessage?.content?.substring(0, 100) + (lastMessage?.content?.length > 100 ? '...' : ''),
|
|
messageCount: messages.length,
|
|
}, 'POST');
|
|
|
|
try {
|
|
const chunks = await this.client.chat.completions.create({
|
|
model,
|
|
messages: messages as any,
|
|
stream: true,
|
|
stream_options: { include_usage: true },
|
|
}, {
|
|
signal: options?.signal,
|
|
});
|
|
|
|
return {
|
|
async *[Symbol.asyncIterator]() {
|
|
try {
|
|
for await (const chunk of chunks) {
|
|
if (options?.signal?.aborted) break;
|
|
yield _transformChunk(chunk);
|
|
}
|
|
const responseTime = Date.now() - startTime;
|
|
logManager.logApiResponse('chat.completions.create', { success: true }, 200, responseTime);
|
|
} catch (error) {
|
|
const responseTime = Date.now() - startTime;
|
|
logManager.logApiResponse('chat.completions.create', { error: error instanceof Error ? error.message : String(error) }, 500, responseTime);
|
|
throw error;
|
|
}
|
|
}
|
|
}
|
|
} catch (error) {
|
|
const responseTime = Date.now() - startTime;
|
|
logManager.logApiResponse('chat.completions.create', { error: error instanceof Error ? error.message : String(error) }, 500, responseTime);
|
|
throw error;
|
|
}
|
|
}
|
|
}
|
|
|