Skip to main content

Documents in Sessions

Individual and group sessions can extract readable text from a document and send it to the advisor as reference material.

Supported files​

The current extractor supports:

  • PDF (.pdf);
  • modern Word documents (.docx);
  • modern PowerPoint files (.pptx);
  • plain text (.txt);
  • Markdown (.md); and
  • CSV (.csv).

It does not support legacy binary .doc or .ppt, images, or archives.

The maximum file size is 15 MB. Very long extracted text is capped before it is sent to the advisor.

Attach in an individual session​

  1. Open an active session you own.
  2. Select Attach a document beside the composer.
  3. Choose one supported file.
  4. Wait for extraction and sending to finish.
  5. Review the new transcript turn, which begins with the uploaded filename.

The turn includes an attachment summary with filename, extracted word count, and original file size.

Attach in a group session​

The workflow is the same. An uploaded document is sent in Ask the assistant mode, even if the text composer had been set to Note to team. This ensures the advisor treats the extracted text as reference material and runs the normal maturity pipeline.

Every group participant can see the resulting transcript content.

Extraction and storage behaviour​

Text extraction occurs in the browser. Navaid:

  1. identifies the format;
  2. extracts text;
  3. cleans unsupported control characters and excessive whitespace;
  4. sends extracted text as a conversation turn; and
  5. records the upload in the company's document history on a best-effort basis.

The deployment can optionally preserve an original-file copy in configured document storage. Company Knowledge indicates whether an original is available to download. Do not assume that the absence of an original removes the extracted text from the conversation.

Scanned and image-only documents​

The current extractor does not perform optical character recognition. A scanned PDF can therefore produce:

No readable text was found in that file.

Create a text-searchable PDF, export an accessible DOCX/PPTX, or paste the relevant text into the conversation.

Large and complex files​

For a better result:

  • remove irrelevant appendices;
  • use a text-searchable export;
  • split unrelated subjects into separate files or turns;
  • identify what the advisor should pay attention to in your next message; and
  • confirm critical facts after upload.

PowerPoint speaker notes can be extracted. PDF and slide processing has protective limits; content after those limits may not reach the advisor.

Sensitive information​

An attachment becomes company-scoped conversation and document-history data:

  • authorised elevated colleagues may be able to view the session;
  • group participants see a group upload;
  • future sessions can use resulting company knowledge;
  • an original file may be retained when storage is configured; and
  • personal data exports can include the owner's session messages.

Remove secrets and unnecessary personal or regulated data before upload. Never use the session as a password, credential, private-key, or service-key store.

Delete or correct uploaded information​

Deleting a document-history entry and deleting a session are separate actions. Removing an original document does not necessarily remove the extracted transcript turn, company facts, or maturity conclusions that were produced earlier.

To correct outdated information:

  1. state the correction clearly in a later conversation;
  2. update Company Knowledge when your role permits;
  3. review provenance/history; and
  4. delete obsolete source records only when appropriate.

Troubleshooting​

“File is too large”​

Reduce it below 15 MB or export the relevant pages/slides into a smaller file.

“Unsupported file type”​

Convert legacy Office files to .docx or .pptx, or export readable text/PDF.

Upload never becomes a message​

Wait for extraction to finish. If an error appears, dismiss it and retry once with a smaller text-based file. Do not navigate away during extraction.

The advisor missed a table or visual​

Extraction is text-focused. Describe the conclusion in a follow-up message or provide a text version of the table. Images and layout are not interpreted as visual evidence.

Last updated on