Merge pull request #113 from solidSpoon/whisper-issue

优化生成字幕的逻辑

Author: solidSpoon

package.json

@@ -1,7 +1,7 @@
 {
     "name": "dash-player",
     "productName": "DashPlayer",
-    "version": "5.1.3",
+    "version": "5.1.4",
     "description": "My Electron application description",
     "main": ".vite/build/main.js",
     "scripts": {
@@ -139,6 +139,7 @@
         "tailwind-merge": "^2.2.1",
         "tailwindcss-animate": "^1.0.7",
         "tencentcloud-sdk-nodejs": "^4.0.764",
+        "ts-error": "^1.0.6",
         "vaul": "^0.9.0",
         "zod": "^3.23.8",
         "zod-to-json-schema": "^3.22.5",

src/backend/errors/CancelByUserError.ts

@@ -0,0 +1,13 @@
+import { ExtendableError } from 'ts-error';
+
+/**
+ * Whisper 相应格式错误
+ */
+export class WhisperResponseFormatError extends ExtendableError {
+}
+
+/**
+ * 任务被用户取消
+ */
+export class CancelByUserError extends ExtendableError {
+}

src/backend/errors/errors.ts

@@ -4,34 +4,9 @@ import RateLimiter from '@/common/utils/RateLimiter';
 import StrUtil from '@/common/utils/str-util';
 import { Cancelable } from '@/common/interfaces';
 import OpenAI from 'openai';
-
-import { z } from 'zod';
 import dpLog from '@/backend/ioc/logger';
-
-const WhisperResponseVerifySchema = z.object({
-    language: z.string(),
-    duration: z.union([z.number(), z.string()]),
-    text: z.string(),
-    segments: z.array(z.object({
-        seek: z.number(),
-        start: z.number(),
-        end: z.number(),
-        text: z.string()
-    }))
-});
-
-export interface WhisperResponse {
-    language: string;
-    duration: number;
-    text: string;
-    offset: number;
-    segments: {
-        seek: number;
-        start: number;
-        end: number;
-        text: string;
-    }[];
-}
+import { WhisperResponseFormatError } from '@/backend/errors/errors';
+import { WhisperResponse, WhisperResponseVerifySchema } from '@/common/types/video-info';
 
 class OpenAiWhisperRequest implements Cancelable {
     private readonly file: string;
@@ -55,33 +30,41 @@ class OpenAiWhisperRequest implements Cancelable {
     public async invoke(): Promise<WhisperResponse> {
         this.cancel();
         await RateLimiter.wait('whisper');
-        this.abortController = new AbortController();
-        const transcription = await this.openAi.audio.transcriptions.create({
-            file: fs.createReadStream(this.file),
-            model: "whisper-1",
-            response_format: "verbose_json",
-            timestamp_granularities: ["segment"]
-        }, {signal: this.abortController.signal});
+        const transcription = await this.doTranscription();
         // 用 zed 校验一下 transcription 是否为 类型 TranscriptionVerbose
         const parseRes = WhisperResponseVerifySchema.safeParse(transcription);
         if (!parseRes.success) {
             // dperror 为什么不匹配
             dpLog.error('Invalid response from OpenAI', parseRes.error.errors);
-            throw new Error('Invalid response from OpenAI');
+            throw new WhisperResponseFormatError();
         }
         return {
             language: transcription.language,
-            duration: Number(transcription.duration),
+            duration: transcription.duration,
             text: transcription.text,
-            offset: 0,
             segments: transcription.segments?.map((seg) => ({
                 seek: seg.seek,
                 start: seg.start,
                 end: seg.end,
                 text: seg.text
-            }))??[]
-        }
+            })) ?? []
+        };
+
+    }
 
+    private async doTranscription() {
+        this.abortController = new AbortController();
+        try {
+            return await this.openAi.audio.transcriptions.create({
+                file: fs.createReadStream(this.file),
+                model: 'whisper-1',
+                response_format: 'verbose_json',
+                timestamp_granularities: ['segment']
+            }, { signal: this.abortController.signal });
+        } catch (error) {
+            dpLog.error(error);
+            throw error;
+        }
     }
 
     public cancel(): void {

src/backend/objs/OpenAiWhisperRequest.ts

@@ -0,0 +1,54 @@
+import { z } from 'zod';
+import fs from 'fs';
+import path from 'path';
+
+/**
+ * 配置文件托管类
+ * @template T 配置类型
+ * @template S Zod Schema 类型
+ */
+export class ConfigTender<T, S extends z.ZodType<T>> {
+    private readonly configPath: string;
+    private readonly schema: S;
+
+    constructor(configPath: string, schema: S, defaultValue?: T) {
+        this.configPath = configPath;
+        this.schema = schema;
+
+        // 确保目录存在
+        const dir = path.dirname(configPath);
+        if (!fs.existsSync(dir)) {
+            fs.mkdirSync(dir, { recursive: true });
+        }
+
+        // 如果文件不存在且提供了默认值，则创建文件
+        if (!fs.existsSync(configPath) && defaultValue) {
+            this.save(defaultValue);
+        }
+    }
+
+    /**
+     * 读取整个配置
+     */
+    get(): T {
+        try {
+            const content = fs.readFileSync(this.configPath, 'utf-8');
+            const parsed = JSON.parse(content);
+            return this.schema.parse(parsed);
+        } catch (error) {
+            throw new Error(`Failed to read config: ${error}`);
+        }
+    }
+
+    /**
+     * 保存整个配置
+     */
+    save(config: T): void {
+        try {
+            const validated = this.schema.parse(config);
+            fs.writeFileSync(this.configPath, JSON.stringify(validated, null, 2));
+        } catch (error) {
+            throw new Error(`Failed to save config: ${error}`);
+        }
+    }
+}

src/backend/objs/config-tender.ts

@@ -1,3 +1,4 @@
+import { VideoInfo } from '@/common/types/video-info';
 
 
 export default interface FfmpegService {
@@ -82,5 +83,10 @@ export default interface FfmpegService {
     }): Promise<string>;
 
     trimVideo(inputPath: string, startTime: number, endTime: number, outputPath: string): Promise<void>;
+
+    /**
+     * Get video information
+     */
+    getVideoInfo(filePath: string): Promise<VideoInfo>;
 }
 

src/backend/services/FfmpegService.ts

@@ -4,12 +4,12 @@ import { DpTask, dpTask, DpTaskState, InsertDpTask } from '@/backend/db/tables/d
 
 import LRUCache from 'lru-cache';
 import TimeUtil from '@/common/utils/TimeUtil';
-import ErrorConstants from '@/common/constants/error-constants';
 import { injectable, postConstruct } from 'inversify';
 import DpTaskService from '@/backend/services/DpTaskService';
 import dpLog from '@/backend/ioc/logger';
 import { Cancelable } from '@/common/interfaces';
-import CancelByUserError from '@/backend/errors/CancelByUserError';
+
+import { CancelByUserError } from '@/backend/errors/errors';
 
 @injectable()
 export default class DpTaskServiceImpl implements DpTaskService {
@@ -188,6 +188,9 @@ export default class DpTaskServiceImpl implements DpTaskService {
     public registerTask(taskId: number, process: Cancelable) {
         const existingProcesses = this.taskMapping.get(taskId) || [];
         this.taskMapping.set(taskId, [...existingProcesses, process]);
+        if (this.cancelQueue.has(taskId)) {
+            process.cancel();
+        }
     }
 
 }

src/backend/services/impl/DpTaskServiceImpl.ts

@@ -9,10 +9,11 @@ import TYPES from '@/backend/ioc/types';
 import FfmpegService from '@/backend/services/FfmpegService';
 import FfmpegTask from '@/backend/objs/FfmpegTask';
 import DpTaskService from '@/backend/services/DpTaskService';
-import CancelByUserError from '@/backend/errors/CancelByUserError';
 import dpLog from '@/backend/ioc/logger';
 import ffmpeg from 'fluent-ffmpeg';
 import LocationService, { ProgramType } from '@/backend/services/LocationService';
+import { VideoInfo } from '@/common/types/video-info';
+import { CancelByUserError } from '@/backend/errors/errors';
 
 @injectable()
 export default class FfmpegServiceImpl implements FfmpegService {
@@ -100,6 +101,34 @@ export default class FfmpegServiceImpl implements FfmpegService {
         });
     }
 
+    /**
+     * 获取视频文件的详细信息
+     */
+    @WaitLock('ffprobe')
+    @logParams()
+    public async getVideoInfo(filePath: string): Promise<VideoInfo> {
+        // 获取文件基本信息
+        const stats = await fs.promises.stat(filePath);
+
+        // 获取ffprobe信息
+        const probeData = await new Promise<any>((resolve, reject) => {
+            ffmpeg.ffprobe(filePath, (err, metadata) => {
+                if (err) reject(err);
+                else resolve(metadata);
+            });
+        });
+
+        return {
+            filename: path.basename(filePath),
+            duration: probeData.format.duration || 0,
+            size: stats.size,
+            modifiedTime: stats.mtimeMs,
+            createdTime: stats.ctimeMs,
+            bitrate: probeData.format.bit_rate ? parseInt(probeData.format.bit_rate) : undefined,
+            videoCodec: probeData.streams.find((s: any) => s.codec_type === 'video')?.codec_name,
+            audioCodec: probeData.streams.find((s: any) => s.codec_type === 'audio')?.codec_name
+        };
+    }
 
     /**
      * 截取视频的缩略图