GenAI eatting image from DOCX

[fadingNA]

• What kind of change does this PR introduce? (Bug fix, feature, docs update, ...)

• 🚀 Feature Allow our AI's to eat images. #1451

Chage from docx2txt.process to extract them manually
Eatting image as Base64 | Table as HTML tag | Text for paragraph

• Why was this change needed? (You can also link to an open issue here)

To using multiple vector store to retrieve correct image as paragraph order instead of zip them by docx2text.process()

• Other information:
• Progress the issue is now when we retrieve need to find the way to convert base64 back for AI to understand image
  

@dartpain let me know if this correct approach or wrong direction

[vercel]

@fadingNA is attempting to deploy a commit to the Arc53 Team on Vercel.

A member of the Team first needs to authorize it.

[fadingNA]

@dartpain this is the progress as we discussed on discord, let me know if this right direction.

• for the text embedding seems work properly except the default document that using 768 dimension but clip using 512 dimension.
• Image maybe I have to resize to make base64 smaller before embed to CLIP model
• Issue here it stuck when I upload docx to server it seems stuck with document that contains images.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
docs-gpt	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Dec 6, 2024 5:52am