Skip to content
OlyvrOlyvrOlyvr Workspace

Import files

Your workspace is the set of documents Olyvr can search and cite. Documents arrive through sources: uploaded files, a folder on your computer, or a connected account.

Supported files

Section titled “Supported files”
Type Extensions
Office .docx, .xlsx, .pptx
PDF .pdf (with a text layer)
Text and data .txt, .md, .csv, .tsv, .html, .htm, .json, .yaml, .yml, .xml, .log
Other .eml (email), .ics (calendar), .py, .js, .ts

Olyvr extracts and indexes the text. Scanned PDFs without a text layer can’t be read yet, because there is no OCR. Password-protected PDFs must be unlocked first.

Add a source

Section titled “Add a source”

  1. Open Workspace from the sidebar.

  2. Choose how to add documents:

    • Files: pick one or more files. Olyvr keeps its own copy of each file.
    • Folder: pick a folder, or enter its full path. Olyvr imports every supported file inside it, including subfolders.
    • Account: connect Google, Microsoft or GitHub. See Connect accounts.

  3. The source appears with its document count once it is ready.

Limits

Section titled “Limits”
  • 30 MB per file (up to 120 MB uncompressed for Office files)
  • 5,000 files per folder; hidden files and links are skipped
  • 1,000 pages per PDF, 100,000 rows per worksheet

Keep sources up to date

Section titled “Keep sources up to date”

Click Refresh on a source to re-import it:

  • Folder sources re-read the folder. Files you deleted from the folder are removed from the workspace.
  • File sources re-extract their stored copies.
  • Account sources collect a fresh snapshot from the provider.

Each document keeps the same identity across refreshes, so existing citations stay valid.

Turn sources off or remove them

Section titled “Turn sources off or remove them”
  • Disable a source to hide its documents from search, reading and citations without deleting anything. Enable it again at any time.
  • Remove a source to delete it from the workspace. Olyvr’s stored copies are deleted shortly afterwards, unless a chat or project still needs them. Files in your own folders are never touched.

Changing your sources marks generated memory notes as out of date, so you can rebuild them from the current workspace.

How search works

Section titled “How search works”

Search is a fast keyword index across every enabled source. It first looks for documents containing all your words, with matches in the file name ranked higher. If nothing matches, it falls back to the most distinctive words in your query. Search does not need a model. It runs entirely on your machine.

The agent uses the same search, then reads the documents that look most relevant before answering.