Scrutium feature guide
Chat with Documents
Turn a document into a focused conversation. Add a supported file to a chat, then ask for a summary, explanation, comparison, or answers grounded in the extracted text.
Start a document chatSupported document formats
Scrutium extracts text from text-based PDFs, DOCX files, XLSX and XLS spreadsheets, CSV files, and common plain-text formats. Spreadsheet content is presented as readable rows and columns for the model to analyze.
- PDF and DOCX
- XLSX, XLS, and CSV
- TXT, Markdown, JSON, and code files
Ask questions about the content
After upload, ask for the key points, a plain-language explanation, or a comparison between sections. The extracted text is added to the model context so responses can refer to the document rather than guessing from its filename.
Know the current limits
Scanned and image-only PDFs do not have OCR yet. Very long files are shortened to fit the model context, spreadsheets are sampled, and legacy DOC files are not supported. Check important answers against the original file.
Example prompts
- “Summarize this document in five key points and cite the relevant section text.”
- “List the assumptions and action items in this file.”
- “Compare the totals in these two spreadsheet columns and explain the difference.”
Common questions
Can Scrutium read a scanned PDF?
Not yet. Text-based PDFs can be extracted, but scanned or image-only PDFs need OCR before their text can be used in chat.
Which spreadsheet files can I upload?
XLSX, XLS, and CSV files are supported. Spreadsheet extraction samples up to 10 sheets and the first 80 rows of each sheet.
Are very large documents processed in full?
No. Extracted text is capped to fit the model context, so a very long document may be truncated. Ask about specific sections or provide a smaller file when completeness matters.