feat: extended-files

[Toddyclipsgg]

Detailed Summary of Document Processing Feature

The implemented feature introduces a robust system for processing documents and files in the application, with a special focus on PDF and DOCX files. Here is a detailed breakdown of the features:

Document Text Extraction

The core component is the implementation of text extractors for different types of documents:

1. DOCX file processing:

• Uses the JSZip library to decompress the contents of the DOCX file
• Parses the internal XML document (word/document.xml) to extract the text
• Uses regular expressions to identify and extract the contents of <w:t> tags
• Reconstructs the text of the document while maintaining basic formatting

2. PDF file processing:

• Implements a simplified approach that does not rely on PDF.js workers
• Parses the raw bytes of the PDF to find text strings
• Looks for common patterns in PDFs such as text enclosed in parentheses
• Identifies uncompressed blocks of text enclosed in /BT and /ET tags
• Extracts TJ and Tj command strings that frequently contain text
• Processes and cleans the extracted text, preserving line breaks line

File Handling in the Interface

The system also implements a complete interface for uploading and manipulating files:

1. File Upload:

• Support for multiple file types (.md, .docx, .pdf, .txt)
• File selection interface with appropriate filters
• Asynchronous processing of selected files
• Validation of file types and maximum size (5MB)

2. Document Preview:

• Display of specific icons per file type
• Formatted preview for PDF and DOCX documents with distinctive icons
• Special handling for text documents with proper formatting
• Informative toast notifications about files attached

3. Manipulation via Drag & Drop and Clipboard:

• Support for dragging and dropping files into the interface
• Capture of pasted files from the clipboard
• Security validation to prevent uploading of scripted files
• Visual feedback during the upload process

4. Integration with the Chat System:

• Adding files to the conversation context
• Processing of documents before sending the message
• Extracting text for analysis by the language model
• State management for attached files

The feature implements a complete flow from uploading to processing and displaying documents, with a focus on user experience and security, avoiding potential problems with malicious or very large files.

Resource created from request: #1412 @leex279

[thecodacus]

i think we should change the order,

Here is some context from files for your reference\n\n---\n ${textFilesInfo}\n---\n${messageContent}

[Toddyclipsgg]

i think we should change the order,

Here is some context from files for your reference\n\n---\n ${textFilesInfo}\n---\n${messageContent}

Can I reverse the order then?

[thecodacus]

yes.. the content at the end gets more emphasis, so the AI will know what to do with all the context, if we put that after the context

[thecodacus]

I just got chance to review the code. looks goof to me.. @leex279 if you can test the functionality and UI then we are good I believe

[leex279]

will test later today

[Toddyclipsgg]

I just got chance to review the code. looks goof to me.. @leex279 if you can test the functionality and UI then we are good I believe

Thanks @thecodacus for reviewing the feature!

[leex279]

Findings

PDF-Files look not working properly. Either it gets an error (maybe to big) or it does not use the pdf content as instruction, instead building a pdf viewer :D
1. Not working at all Example: https://www.cmu.edu/swartz-center-for-entrepreneurship/assets/creating-an-effective-landing-page.pdf

2. Building wrong app:
Unbenanntes Dokument.pdf

3. Styling could be improved (margin between filebox and chat):

[Toddyclipsgg]

Can you please test it for me now? @leex279

[Toddyclipsgg]

@thecodacus I ended up deleting the commits of the feature by accident if you can see if I didn't do anything wrong when I recovered the commits please!

[leex279]

@thecodacus please take a look. I dont know :D ... let me know when I can test/review again :)