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
|
import os from "os";
import { eq } from "drizzle-orm";
import { DequeuedJob, Runner } from "liteque";
import PDFParser from "pdf2json";
import { createWorker } from "tesseract.js";
import type { AssetPreprocessingRequest } from "@hoarder/shared/queues";
import { db } from "@hoarder/db";
import { bookmarkAssets, bookmarks } from "@hoarder/db/schema";
import { readAsset } from "@hoarder/shared/assetdb";
import serverConfig from "@hoarder/shared/config";
import logger from "@hoarder/shared/logger";
import {
AssetPreprocessingQueue,
OpenAIQueue,
triggerSearchReindex,
} from "@hoarder/shared/queues";
export class AssetPreprocessingWorker {
static build() {
logger.info("Starting asset preprocessing worker ...");
const worker = new Runner<AssetPreprocessingRequest>(
AssetPreprocessingQueue,
{
run: run,
onComplete: async (job) => {
const jobId = job.id;
logger.info(`[assetPreprocessing][${jobId}] Completed successfully`);
return Promise.resolve();
},
onError: async (job) => {
const jobId = job.id;
logger.error(
`[assetPreprocessing][${jobId}] Asset preprocessing failed: ${job.error}\n${job.error.stack}`,
);
return Promise.resolve();
},
},
{
concurrency: 1,
pollIntervalMs: 1000,
timeoutSecs: 30,
},
);
return worker;
}
}
async function readImageText(buffer: Buffer) {
if (serverConfig.ocr.langs.length == 1 && serverConfig.ocr.langs[0] == "") {
return null;
}
const worker = await createWorker(serverConfig.ocr.langs, undefined, {
cachePath: serverConfig.ocr.cacheDir ?? os.tmpdir(),
});
try {
const ret = await worker.recognize(buffer);
if (ret.data.confidence <= serverConfig.ocr.confidenceThreshold) {
return null;
}
return ret.data.text;
} finally {
await worker.terminate();
}
}
async function readPDFText(buffer: Buffer): Promise<{
text: string;
metadata: Record<string, string>;
}> {
return new Promise((resolve, reject) => {
// Need raw text flag represents as number (1), reference : https://github.com/modesty/pdf2json/issues/76#issuecomment-236569265
const pdfParser = new PDFParser(null, 1);
pdfParser.on("pdfParser_dataError", reject);
pdfParser.on("pdfParser_dataReady", (pdfData) => {
resolve({
// The type isn't set correctly, reference : https://github.com/modesty/pdf2json/issues/327
// eslint-disable-next-line
text: (pdfParser as any).getRawTextContent(),
metadata: pdfData.Meta,
});
});
pdfParser.parseBuffer(buffer);
});
}
async function preprocessImage(
jobId: string,
asset: Buffer,
): Promise<{ content: string; metadata: string | null } | undefined> {
let imageText = null;
try {
imageText = await readImageText(asset);
} catch (e) {
logger.error(
`[assetPreprocessing][${jobId}] Failed to read image text: ${e}`,
);
}
if (!imageText) {
return undefined;
}
logger.info(
`[assetPreprocessing][${jobId}] Extracted ${imageText.length} characters from image.`,
);
return { content: imageText, metadata: null };
}
async function preProcessPDF(
jobId: string,
asset: Buffer,
): Promise<{ content: string; metadata: string | null } | undefined> {
const pdfParse = await readPDFText(asset);
if (!pdfParse?.text) {
throw new Error(
`[assetPreprocessing][${jobId}] PDF text is empty. Please make sure that the PDF includes text and not just images.`,
);
}
logger.info(
`[assetPreprocessing][${jobId}] Extracted ${pdfParse.text.length} characters from pdf.`,
);
return {
content: pdfParse.text,
metadata: pdfParse.metadata ? JSON.stringify(pdfParse.metadata) : null,
};
}
async function run(req: DequeuedJob<AssetPreprocessingRequest>) {
const jobId = req.id;
const bookmarkId = req.data.bookmarkId;
const bookmark = await db.query.bookmarks.findFirst({
where: eq(bookmarks.id, bookmarkId),
with: {
asset: true,
},
});
logger.info(
`[assetPreprocessing][${jobId}] Starting an asset preprocessing job for bookmark with id "${bookmarkId}"`,
);
if (!bookmark) {
throw new Error(`[assetPreprocessing][${jobId}] Bookmark not found`);
}
if (!bookmark.asset) {
throw new Error(
`[assetPreprocessing][${jobId}] Bookmark is not an asset (not an image or pdf)`,
);
}
const { asset } = await readAsset({
userId: bookmark.userId,
assetId: bookmark.asset.assetId,
});
if (!asset) {
throw new Error(
`[assetPreprocessing][${jobId}] AssetId ${bookmark.asset.assetId} for bookmark ${bookmarkId} not found`,
);
}
let result: { content: string; metadata: string | null } | undefined =
undefined;
switch (bookmark.asset.assetType) {
case "image":
result = await preprocessImage(jobId, asset);
break;
case "pdf":
result = await preProcessPDF(jobId, asset);
break;
default:
throw new Error(
`[assetPreprocessing][${jobId}] Unsupported bookmark type`,
);
}
if (result) {
await db
.update(bookmarkAssets)
.set({
content: result.content,
metadata: result.metadata,
})
.where(eq(bookmarkAssets.id, bookmarkId));
}
await OpenAIQueue.enqueue({
bookmarkId,
});
// Update the search index
await triggerSearchReindex(bookmarkId);
}
|
