SideNote ProAsk AI about PDFs, Word, Excel, PowerPoint and email files
Drop a specification, a spreadsheet, a slide deck or a saved email onto SideNote Pro and ask about it. Reading happens on your PC, the original file is never uploaded, and page, sheet and slide labels survive so an answer can point at where it found something.
Supported formats and limits
| Property | Value |
|---|---|
| Documents | PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx) and saved emails (.eml) |
| Text, data and code | TXT and Markdown (.md, .markdown), plus CSV, JSON, XML, YAML, logs, configuration files and source code |
| Not accepted | Older and macro-enabled Office files (.doc, .xls, .ppt, .xlsm, .pptm) and Outlook .msg files |
| Documents per request | Up to 5 |
| Maximum input file size | 20 MB |
| Maximum extracted text | 2,000,000 characters |
| Attachment methods | Attach documents in the composer, or drag and drop |
| Oversized context | Rejected with a clear message, never silently truncated |
SideNote Pro rejects a request that exceeds the context window instead of silently truncating the document. Otherwise, a model could answer confidently about a document it only saw part of. You can shorten the input or raise the context window knowingly.
Extraction happens locally
Attached files are read on your machine and shown as separate attachment chips before you send anything. Attaching a file never starts Office or any other program, and nothing is calculated, fetched or sent while it is read. Large files are read in the background so the SideNote Pro window stays responsive; the composer shows which file it is reading, and Send waits for it to finish. What enters the request is extracted text, delimited as untrusted reference material inside the ordinary Chat Completions call.
A password-protected Office file is refused, and so is an older format renamed to look like a newer one. A damaged file says it could not be read, and none leaves a partial attachment behind. Word, PDF, Excel and PowerPoint files built to take a great deal of memory are refused as too large or nested too deeply to read safely.
- Text extraction preserves page markers, so a response can attribute a point to a page.
- Word
- Document text, lists and tables are read without running macros, loading external resources or pulling in embedded objects.
- Excel
- Each sheet comes in the workbook's own order as a table with column letters and row numbers, so every value keeps its cell address. Hidden sheets are marked as hidden.
- PowerPoint
- Each slide comes in order with its text, its tables and its speaker notes. Hidden slides are marked as hidden.
- The subject, sender, recipients and date come first, then the message text.
- Text, data and code
- Read directly, exactly as written.
What Excel, PowerPoint and email files include
| Format | Read | Not read |
|---|---|---|
Excel (.xlsx) | Cell values, text and dates. A formula shows the result Excel last saved, and a formula with no saved result is shown as its formula. | Charts, chart sheets and number formats other than dates, so 15% reads as 0.15. Nothing is recalculated. |
PowerPoint (.pptx) | Text in each slide's shapes and groups, tables, and speaker notes. | Pictures, charts, SmartArt, media, embedded objects, slide masters and layouts. |
Email (.eml) | Subject, From, To, Cc and Date, then the message's plain text, or its HTML turned into text. | Attachments inside the email, which are listed by name. Encrypted messages are refused. |
Scanned PDFs and pages as images
SideNote Pro does not run text recognition (OCR). A scanned PDF holds pictures of text, so when a PDF has no readable text, SideNote Pro offers to read its pages as images instead. You choose up to five pages, they are drawn on your PC, and they are sent as images when you press Send. They need a model that accepts images. A hosted provider may charge for image input; a local model has no API charge.
The PDF itself is still never uploaded: only the pages you chose travel, as images. Asking AI about scanned PDF pages walks through choosing pages, the image limits and what to ask.
- A PDF with no text at all
- Attaching it asks whether to read its pages as images, and opens the page selector if you continue.
- A PDF with some pages without text
- It is attached as text as usual, and its chip says how many pages have no text, since they may be scanned or blank.
- Any PDF
- Read pages as images…, under the chip's More actions, turns chosen pages into images, for a diagram or a chart text cannot carry. The pages you choose replace the whole text attachment.
The privacy boundary
SideNote Pro reads documents on your machine. What it extracts reaches your AI provider when you press Send.
Stays on your machine
The original file. It is read, parsed and closed. SideNote Pro never transmits the file itself, never opens it in Office, and never opens a network connection while reading it.
Goes to your provider
The extracted text, or the PDF pages you chose as images, once you press Send. If the document is confidential, consider how your provider handles it.
Extracted text and document metadata are stored inside the conversation archive, so a saved thread still makes sense later even if you moved or deleted the original file. Enabling history protection applies DPAPI to document contents along with everything else. Markdown export lists attachment names without dumping their extracted bodies.
Getting good answers from documents
- Ask for structure, not just a summary
- Extract every dated commitment and who owns it beats summarize this.
- Attach two documents to compare them
- Two proposals, two contract versions, a specification and its implementation notes. Comparison is where the five-document limit earns its keep.
- Point at the part you mean
- With page, sheet and slide labels preserved, what does the section on data retention actually require or which rows on the Q3 sheet are over budget works better than making the model guess.
- Attach a workbook when the whole sheet matters
- A selection or a screenshot shows part of a sheet; an attached workbook gives the model every sheet with its cell addresses. Using AI beside Excel compares the three.
- Watch the context window
- A large file plus a long conversation can exceed it. Managing conversations explains compaction and how to raise it.
- Verify before you rely on it
- For anything with legal or financial consequence, an AI reading of a document is a first pass, not an authority.
For a walkthrough with concrete prompts, see asking AI about documents. If you find yourself attaching the same set of files repeatedly, local folder context is the better tool.