fix(build): consolidate pdf parsing dependencies, remove extraneous html deps (#1212)

* fix(build): consolidate pdf parsing dependencies, remove extraneous html deps

* add types

Author: waleedlatif1

apps/sim/app/api/files/parse/route.ts

@@ -553,22 +553,11 @@ function handleGenericBuffer(
  */
 async function parseBufferAsPdf(buffer: Buffer) {
   try {
-    try {
-      const { PdfParser } = await import('@/lib/file-parsers/pdf-parser')
-      const parser = new PdfParser()
-      logger.info('Using main PDF parser for buffer')
-
-      if (parser.parseBuffer) {
-        return await parser.parseBuffer(buffer)
-      }
-      throw new Error('PDF parser does not support buffer parsing')
-    } catch (error) {
-      logger.warn('Main PDF parser failed, using raw parser for buffer:', error)
-      const { RawPdfParser } = await import('@/lib/file-parsers/raw-pdf-parser')
-      const rawParser = new RawPdfParser()
+    const { PdfParser } = await import('@/lib/file-parsers/pdf-parser')
+    const parser = new PdfParser()
+    logger.info('Using main PDF parser for buffer')
 
-      return await rawParser.parseBuffer(buffer)
-    }
+    return await parser.parseBuffer(buffer)
   } catch (error) {
     throw new Error(`PDF parsing failed: ${(error as Error).message}`)
   }

apps/sim/lib/file-parsers/index.test.ts

@@ -141,17 +141,6 @@ describe('File Parsers', () => {
       })),
     }))
 
-    vi.doMock('@/lib/file-parsers/raw-pdf-parser', () => ({
-      RawPdfParser: vi.fn().mockImplementation(() => ({
-        parseFile: vi.fn().mockResolvedValue({
-          content: 'Raw parsed PDF content',
-          metadata: {
-            pageCount: 3,
-          },
-        }),
-      })),
-    }))
-
     vi.doMock('@/lib/file-parsers/txt-parser', () => ({
       TxtParser: vi.fn().mockImplementation(() => ({
         parseFile: mockTxtParseFile,

apps/sim/lib/file-parsers/index.ts

@@ -1,7 +1,5 @@
 import { existsSync } from 'fs'
-import { readFile } from 'fs/promises'
 import path from 'path'
-import { RawPdfParser } from '@/lib/file-parsers/raw-pdf-parser'
 import type { FileParseResult, FileParser, SupportedFileType } from '@/lib/file-parsers/types'
 import { createLogger } from '@/lib/logs/console/logger'
 
@@ -18,42 +16,12 @@ function getParserInstances(): Record<string, FileParser> {
 
     try {
       try {
-        logger.info('Attempting to load PDF parser...')
-        try {
-          const { PdfParser } = require('@/lib/file-parsers/pdf-parser')
-          parserInstances.pdf = new PdfParser()
-          logger.info('PDF parser loaded successfully')
-        } catch (pdfLibError) {
-          logger.error('Failed to load primary PDF parser:', pdfLibError)
-          logger.info('Falling back to raw PDF parser')
-          parserInstances.pdf = new RawPdfParser()
-          logger.info('Raw PDF parser loaded successfully')
-        }
+        logger.info('Loading PDF parser...')
+        const { PdfParser } = require('@/lib/file-parsers/pdf-parser')
+        parserInstances.pdf = new PdfParser()
+        logger.info('PDF parser loaded successfully')
       } catch (error) {
-        logger.error('Failed to load any PDF parser:', error)
-        parserInstances.pdf = {
-          async parseFile(filePath: string): Promise<FileParseResult> {
-            const buffer = await readFile(filePath)
-            return {
-              content: `PDF parsing is not available. File size: ${buffer.length} bytes`,
-              metadata: {
-                info: { Error: 'PDF parsing unavailable' },
-                pageCount: 0,
-                version: 'unknown',
-              },
-            }
-          },
-          async parseBuffer(buffer: Buffer): Promise<FileParseResult> {
-            return {
-              content: `PDF parsing is not available. File size: ${buffer.length} bytes`,
-              metadata: {
-                info: { Error: 'PDF parsing unavailable' },
-                pageCount: 0,
-                version: 'unknown',
-              },
-            }
-          },
-        }
+        logger.error('Failed to load PDF parser:', error)
       }
 
       try {

apps/sim/lib/file-parsers/pdf-parser.ts

@@ -1,11 +1,9 @@
 import { readFile } from 'fs/promises'
-import { PDFDocument } from 'pdf-lib'
+import pdfParse from 'pdf-parse'
 import type { FileParseResult, FileParser } from '@/lib/file-parsers/types'
 import { createLogger } from '@/lib/logs/console/logger'
-import { RawPdfParser } from './raw-pdf-parser'
 
 const logger = createLogger('PdfParser')
-const rawPdfParser = new RawPdfParser()
 
 export class PdfParser implements FileParser {
   async parseFile(filePath: string): Promise<FileParseResult> {
@@ -31,68 +29,23 @@ export class PdfParser implements FileParser {
     try {
       logger.info('Starting to parse buffer, size:', dataBuffer.length)
 
-      try {
-        logger.info('Attempting to parse with pdf-lib library...')
-
-        logger.info('Starting PDF parsing...')
-        const pdfDoc = await PDFDocument.load(dataBuffer)
-        const pages = pdfDoc.getPages()
-        const pageCount = pages.length
-
-        logger.info('PDF parsed successfully with pdf-lib, pages:', pageCount)
-
-        const metadata: Record<string, any> = {
-          pageCount,
-        }
-
-        try {
-          const title = pdfDoc.getTitle()
-          const author = pdfDoc.getAuthor()
-          const subject = pdfDoc.getSubject()
-          const creator = pdfDoc.getCreator()
-          const producer = pdfDoc.getProducer()
-          const creationDate = pdfDoc.getCreationDate()
-          const modificationDate = pdfDoc.getModificationDate()
-
-          if (title) metadata.title = title
-          if (author) metadata.author = author
-          if (subject) metadata.subject = subject
-          if (creator) metadata.creator = creator
-          if (producer) metadata.producer = producer
-          if (creationDate) metadata.creationDate = creationDate.toISOString()
-          if (modificationDate) metadata.modificationDate = modificationDate.toISOString()
-        } catch (metadataError) {
-          logger.warn('Could not extract PDF metadata:', metadataError)
-        }
-
-        logger.info(
-          'pdf-lib loaded successfully, but text extraction requires fallback to raw parser'
-        )
-        const rawResult = await rawPdfParser.parseBuffer(dataBuffer)
-
-        return {
-          content: rawResult.content,
-          metadata: {
-            ...rawResult.metadata,
-            ...metadata,
-            source: 'pdf-lib + raw-parser',
-          },
-        }
-      } catch (pdfLibError: unknown) {
-        logger.error('PDF-lib library failed:', pdfLibError)
-
-        logger.info('Falling back to raw PDF parser...')
-        const rawResult = await rawPdfParser.parseBuffer(dataBuffer)
-
-        return {
-          ...rawResult,
-          metadata: {
-            ...rawResult.metadata,
-            fallback: true,
-            source: 'raw-parser-only',
-            error: (pdfLibError as Error).message || 'Unknown error',
-          },
-        }
+      const pdfData = await pdfParse(dataBuffer)
+
+      logger.info(
+        'PDF parsed successfully, pages:',
+        pdfData.numpages,
+        'text length:',
+        pdfData.text.length
+      )
+
+      return {
+        content: pdfData.text,
+        metadata: {
+          pageCount: pdfData.numpages,
+          info: pdfData.info,
+          version: pdfData.version,
+          source: 'pdf-parse',
+        },
       }
     } catch (error) {
       logger.error('Error parsing buffer:', error)