Haruno Knowledge base

How Haruno reads your files

What Haruno accepts as an attachment, which documents it reads for summaries and how far, and what it never opens or runs.

  • Mac appAvailable
  • Windows appAvailable

Haruno reads the files you give it, but it is careful with them. Attaching a file never makes it run. This page explains what Haruno lets in, what it reads, and what it will never open.

It looks at what a file really is

Haruno does not trust a file name alone. A file called report.pdf that is really a program is treated as a program. A file called notes.txt that is really a zip file does not pass for a text file. The same rule applies everywhere Haruno takes in a file.

What you can attach

You can attach almost anything. Haruno turns away only two kinds of file:

  • Programs, such as .exe, .dll, .msi and .jar.
  • Shortcuts and launchers, such as .lnk, .url and .rdp.

It says: "Programs and shortcuts can’t be attached." The file gets no chip.

Scripts, documents with macros, PDFs with active content and zip files can all be attached, because reading them is the point. Haruno never opens, previews or imports them as something that runs.

If you attach a folder, Haruno takes it whole. When it copies the folder for the chat, a program or shortcut inside it is left out and the rest is used, without a message.

Where each kind of file is allowed

WhereWhat is allowed
Attach to a chatEverything except programs and shortcuts
Open on the Files page or on a file card in a chatOrdinary pictures, text, PDFs, Office and OpenDocument files and zip-type archives, but only when the contents match the name. Scripts, documents with macros, disk images, PDFs with actions and unusual file types are not opened. On a file card, a refused Open says "File could not be opened." A web page, Word, PDF or Excel file also asks Open this file? first. See Working with messages.
The Preview panelDocuments. A program, shortcut, disk image, macro or active document, or a file whose contents do not match its name shows "That file type cannot be shown here."
A file added to a connector page, such as the PortfolioData files and archives whose contents match their name. Programs, shortcuts, disk images, macro and active documents are refused.

Summaries of Word, Excel, PowerPoint and PDF files

When you select a file on the Files page, Haruno writes a short Summary of it.

The Summary panel under the file list.
The Summary panel on the Files page, showing the selected file's name, type, size and dates and a one-line description of it.The Summary panel on the Files page, showing the selected file's name, type, size and dates and a one-line description of it.

For .docx, .xlsx, .pptx and .pdf files it first reads the text. It does this in an isolated space that has no network and no access to your other files. Older .doc, .xls and .ppt files are not read this way.

What you may notice:

  • Files over 64 MB are not read.
  • A PDF or a slide deck over 500 pages or slides is not read.
  • A workbook is read up to 64 sheets and 200,000 rows.
  • Only about 256 KB of text is taken from one file.
  • Macros, links and embedded objects are never run or followed. A formula shows the value that was last saved.
  • The first time a file is read can take a couple of seconds. The result is remembered while the file stays the same.

If a file cannot be read, the summary says so and goes by the file's details alone. Sometimes the panel just says "No summary available."

The summary is written by Haruno from an excerpt of the file's text.

Every file is also scanned for malware

Files you attach, files that are summarised and files you import are checked by the antivirus on the PC. See Scanning attachments for malware.

Some honest limits

  • Haruno refuses a few legitimate files. A password-protected Office file, or an old .doc that mentions macros in its text, is treated like a macro document. It cannot be previewed, imported or opened from Haruno, but it can still be attached.
  • A PDF with an embedded file, such as some e-invoices, counts as active and is refused for opening, preview and import. It can still be attached.
  • Attaching a file does not give Haruno permission to run anything. Running a command is a separate decision, governed by Haruno's permission rules.

Last updated · PR #2944 · View this version on GitHub

History (1 version)
  1. Internal knowledge base for staff: site, reader keys, assistant, 197 pages and the PR rule (#2944)
    PR #2944Commit a2a7836The page at this version
    What changed
    @@ -0,0 +1,72 @@
+---
+title: How Haruno reads your files
+summary: What Haruno accepts as an attachment, which documents it reads for summaries and how far, and what it never opens or runs.
+feature: chat-file-reading
+covers:
+  - windows-client/docs/file-ingestion.md
+  - windows-client/docs/document-processing.md
+  - windows-client/Haruno.Desktop.Core/Security/**
+  - IntegrationPlatform/Haruno.DocumentExtraction/**
+  - windows-client/Haruno.Desktop.Services/Files/SandboxedDocumentTextExtractor.cs
+  - windows-client/Haruno.Desktop.Services/Files/FileSummaryService.cs
+  - windows-client/haruno-client/Services/FileLaunchService.cs
+---
+
+Haruno reads the files you give it, but it is careful with them. Attaching a file never makes it run. This page explains what Haruno lets in, what it reads, and what it will never open.
+
+## It looks at what a file really is
+
+Haruno does not trust a file name alone. A file called `report.pdf` that is really a program is treated as a program. A file called `notes.txt` that is really a zip file does not pass for a text file. The same rule applies everywhere Haruno takes in a file.
+
+## What you can attach
+
+You can attach almost anything. Haruno turns away only two kinds of file:
+
+- **Programs**, such as `.exe`, `.dll`, `.msi` and `.jar`.
+- **Shortcuts and launchers**, such as `.lnk`, `.url` and `.rdp`.
+
+It says: "Programs and shortcuts can’t be attached." The file gets no chip.
+
+Scripts, documents with macros, PDFs with active content and zip files can all be attached, because reading them is the point. Haruno never opens, previews or imports them as something that runs.
+
+If you attach a folder, Haruno takes it whole. When it copies the folder for the chat, a program or shortcut inside it is left out and the rest is used, without a message.
+
+## Where each kind of file is allowed
+
+| Where | What is allowed |
+|---|---|
+| Attach to a chat | Everything except programs and shortcuts |
+| **Open** on the Files page or on a file card in a chat | Ordinary pictures, text, PDFs, Office and OpenDocument files and zip-type archives, but only when the contents match the name. Scripts, documents with macros, disk images, PDFs with actions and unusual file types are not opened. On a file card, a refused Open says "File could not be opened." A web page, Word, PDF or Excel file also asks **Open this file?** first. See [Working with messages](/windows/chat/working-with-messages/#links-pictures-and-files-in-an-answer). |
+| The **Preview** panel | Documents. A program, shortcut, disk image, macro or active document, or a file whose contents do not match its name shows "That file type cannot be shown here." |
+| A file added to a connector page, such as the Portfolio | Data files and archives whose contents match their name. Programs, shortcuts, disk images, macro and active documents are refused. |
+
+## Summaries of Word, Excel, PowerPoint and PDF files
+
+When you select a file on the **Files** page, Haruno writes a short **Summary** of it.
+
+<Shot src="/m/69d21fbda0e7-windows-chat-file-summary-light.png" dark="/m/a559fe78ca46-windows-chat-file-summary-dark.png" alt="The Summary panel on the Files page, showing the selected file's name, type, size and dates and a one-line description of it." caption="The Summary panel under the file list." />
+
+For `.docx`, `.xlsx`, `.pptx` and `.pdf` files it first reads the text. It does this in an isolated space that has no network and no access to your other files. Older `.doc`, `.xls` and `.ppt` files are not read this way.
+
+What you may notice:
+
+- Files over 64 MB are not read.
+- A PDF or a slide deck over 500 pages or slides is not read.
+- A workbook is read up to 64 sheets and 200,000 rows.
+- Only about 256 KB of text is taken from one file.
+- Macros, links and embedded objects are never run or followed. A formula shows the value that was last saved.
+- The first time a file is read can take a couple of seconds. The result is remembered while the file stays the same.
+
+If a file cannot be read, the summary says so and goes by the file's details alone. Sometimes the panel just says "No summary available."
+
+The summary is written by Haruno from an excerpt of the file's text.
+
+<Callout variant="info" title="Every file is also scanned for malware">
+Files you attach, files that are summarised and files you import are checked by the antivirus on the PC. See [Scanning attachments for malware](/windows/chat/malware-scanning/).
+</Callout>
+
+## Some honest limits
+
+- Haruno refuses a few legitimate files. A password-protected Office file, or an old `.doc` that mentions macros in its text, is treated like a macro document. It cannot be previewed, imported or opened from Haruno, but it can still be attached.
+- A PDF with an embedded file, such as some e-invoices, counts as active and is refused for opening, preview and import. It can still be attached.
+- Attaching a file does not give Haruno permission to run anything. Running a command is a separate decision, governed by Haruno's permission rules.