Skip to main content

^ Document Reading Tools

Careti's AI can directly read various document files just by providing the path. Users don't need to attach files manually.

Difference from Cline

FeatureClineCareti
PDF ReadingOnly when user attachesAI reads directly by path
DOCX/XLSX ReadingOnly when user attachesAI reads directly by path
Hangul (HWPX)Not supportedSupported
Hangul 5.0 (HWP)Not supportedSupported
PowerPoint (PPTX)Not supportedSupported
Jupyter NotebookNot supportedSupported

Supported Formats

FormatExtensionDescription
PDF.pdfText extraction
Word.docxModern Word documents
Excel.xlsxModern Excel spreadsheets
PowerPoint.pptxSlide text extraction
Hangul (Modern).hwpxHangul 2014 and later
Hangul (Legacy).hwpHangul 5.0 ~ 2010
Jupyter.ipynbNotebook cell contents

Unsupported Formats

Legacy binary formats are not supported:

FormatExtensionAlternative
Word 97-2003.docConvert to .docx
Excel 97-2003.xlsConvert to .xlsx
PowerPoint 97-2003.pptConvert to .pptx

Conversion tools: LibreOffice, Google Docs, Microsoft Office

How to Use

Example 1: Analyzing Specification Documents

User: Analyze docs/spec.pdf
AI: [Uses Document Reading Tool] → Extracts PDF content → Provides analysis results

Example 2: Reading Hangul Documents

User: Summarize the contents of contract.hwp
AI: [Uses Document Reading Tool] → Extracts HWP text → Provides summary

Example 3: Analyzing Excel Data

User: Analyze the data in sales-report.xlsx
AI: [Uses Document Reading Tool] → Extracts spreadsheet data → Provides analysis results

Security

Path Protection

  • Path normalization prevents directory traversal attacks
  • Files inside workspace are auto-approved
  • Files outside workspace require user approval

File Size Limit

  • Maximum 50MB
  • Clear error message when exceeded

Configuration

Document reading tools are enabled by default. Since it's a read-only operation, there's no separate toggle setting.

Known Limitations

  1. Images/Charts Not Included: Images and charts in documents are not converted to text
  2. Complex Layouts: Tables and multi-column layouts are converted to simple text
  3. Original Formatting Lost: Font, color, and other formatting information is not extracted