Supported Documents

This page lists the types of documents that are supported by the document processing orchestrator.

Document

1. DOCX

2. PDF

3. PPTX

4. XLSX

5. Google Docs (from URL)

6. Google Slides (from URL)

7. Google Spreadsheet (from URL)

Plain Text

1. JSON

2. CSV

3. HTML

4. Java

5. JavaScript (JS)

6. JSX

7. Log files (.log)

8. Markdown

9. Python

10. TypeScript (TS)

11. TSX

12. Plain text files (.txt)

URL

1. Any public URL that is not behind any protection, e.g. IP block and/or anti-bot measures.

2. YouTube URLs - Will be ingested to text; may sometimes fail due to limitations from Google.

Image

These will be ingested as text.

1. HEIC

2. HEIF

3. JPEG (.jpg/.jpeg)

4. PNG

5. WEBP

6. TIFF

Audio

These will be ingested as text.

1. FLAC

2. MP3

3. OGG

4. WAV

Video

These will be ingested as text.

1. MP4

2. MPEG

3. MOV

4. AVI

5. MKV

6. WEBM

7. FLV

8. WMV

9. 3GP

10. OGV

11. ASF

12. MP2T

13. OGG

Output

Document processing orchestrator can save into the following data stores:

1. Vector database

2. Tabular database

3. Knowledge Graph

Limitations

These limitations are not planned to be supported:

1. PDF: Cannot extract advanced math equations.

2. DOCX: Cannot extract math equations.

3. URL:

   1. Cannot bypass protected URLs.

      1. Might be solved using FirecrawlDownloader (leverages Firecrawl).

   2. Cannot access social media (e.g. Facebook, Instagram, X, TikTok).

   3. Cannot get a specific part from HTML.

      1. Customization available by extending the gllm-docproc library.

4. Cannot process executable or package files (e.g. DMG, EXE, GZ, TAR, ZIP).

5. Cannot process files with proprietary exstensions (e.g. AI, PSD, DLL).

6. Cannot crawl or scrape URLs periodically.

   1. Projects are responsible to manage their own scheduler or cron-jobs.