Files in datarooms

Files#

Upload files to the Dataroom by clicking on the Upload Files button.

Supported File Types#

AITEZA supports a wide variety of file formats for document processing. Files are automatically processed and indexed for retrieval.

Documents

• PDF (.pdf): Portable Document Format, including scanned documents
• Word (.doc, .docx): Microsoft Word documents (legacy and modern)
• PowerPoint (.ppt, .pptx): Microsoft PowerPoint presentations
• Excel (.xls, .xlsx): Microsoft Excel spreadsheets
• OpenDocument (.odt, .ods, .odp): LibreOffice and OpenOffice text documents, spreadsheets, and presentations
• RTF (.rtf): Rich Text Format
• Plain Text (.txt): Plain text files
• Markdown (.md): Markdown formatted text
• CSV (.csv): Comma-separated values
• HTML/XML (.html, .htm, .xml): Web and markup files

Images

• JPEG (.jpg, .jpeg): Common image format
• PNG (.png): Portable Network Graphics
• TIFF (.tiff, .tif): Tagged Image File Format
• BMP (.bmp): Bitmap images
• GIF (.gif): Graphics Interchange Format

Audio (with automatic transcription)

• MP3 (.mp3): MPEG Audio Layer 3
• WAV (.wav): Waveform Audio File Format
• FLAC (.flac): Free Lossless Audio Codec
• AAC (.aac, .m4a): Advanced Audio Coding
• OGG (.ogg): Ogg Vorbis audio
• WebM (.webm): WebM audio
• WMA (.wma): Windows Media Audio
• AIFF (.aiff): Audio Interchange File Format

Email

• Outlook (.msg): Microsoft Outlook messages
• EML (.eml): Standard email format (RFC 822)
• MBOX (.mbox): Mailbox format for multiple emails

Archives

• ZIP (.zip): Common archive format
• TAR (.tar): Tape archive format
• 7-Zip (.7z): 7-Zip compressed archive
• GZIP (.gz): Gzip compressed file
• BZIP2 (.bz2): Bzip2 compressed file
• XZ (.xz): XZ compressed file

Files over 100 MB are stored but not processed into searchable content.

Selecting Multiple Files#

The file list offers a select-all option, letting you select every file at once instead of clicking through them individually. This speeds up bulk actions on many files, such as moving several files to another dataroom at once or retrying embeddings for multiple files together.

File Content Viewer#

When you open a file from the file list, the content viewer displays the extracted markdown representation. For PDFs, the original document is shown alongside the markdown. Word, Excel, PowerPoint, and supported OpenDocument files can also be opened directly as a PDF preview, so you can quickly review their content without downloading the file or opening it in a separate program. The viewer provides several tools:

• Version History (Versionsverlauf): View all versions of the file with type indicators and notes.
• Copy MD (MD kopieren): Copy the extracted markdown content to your clipboard.
• Download: Download the original source file or the extracted markdown file.
• Edit (Bearbeiten): Open the built-in markdown editor to modify file content directly.
• Search Highlighting: Search terms are highlighted within the document content for quick navigation.

File Versioning#

Files in datarooms have full version history. Every change, whether an initial upload, a manual edit, a reprocessing, or a file replacement, creates a new version that is tracked and can be restored.

Editing File Content#

You can edit file content directly using the built-in rich-text markdown editor. Each edit creates a new version automatically, preserving the full history.

Uploading a New File Version#

You can upload a new file to replace the source document while preserving the full version history. This is useful when you have an updated version of a document but want to keep the previous versions accessible.

Audio Files and Transcripts#

Audio files are transcribed into text instead of being processed like documents. While a recording is being transcribed, the file list shows the status Transcribing (Transkribieren).

When you open a transcribed recording, the content viewer shows the transcript instead of the extracted markdown:

• Player: Play the recording, jump to any position and change the speed (1×, 1.25×, 1.5×, 2×). The passage being spoken is highlighted and kept in view.
• Passages with times: Every passage has its own start time. Click a time or the text of a passage to play the recording from there.
• Speakers: Where the transcription model can tell speakers apart, passages are labelled Speaker 1 (Sprecher 1), Speaker 2 and so on. The name is shown whenever the speaker changes.
• Search (Im Transkript suchen): Search terms are highlighted in the transcript. Use Enter or the arrow buttons to move between the results.
• Citations: If you open a recording from a source in a chat answer, the quoted words are highlighted and scrolled into view. A citation of an older version shows that version’s transcript; the player is hidden there because the recording may have been replaced since.

Moving Files to Another Dataroom#

Owners and Editors can move a file to a different Dataroom they have access to. Moving a file removes it from the original Dataroom and re-indexes it in the target Dataroom; future retrieval will use the target Dataroom only.

Reprocessing Files#

You can reprocess existing files to update their content and embeddings using different processing pipelines. This is useful when:

• Processing algorithms have been updated.
• The initial processing did not extract content correctly (e.g., scanned documents).
• You want to try a different processing model for better results.

Processing Pipelines#

AITEZA offers four processing pipelines. Select the level that matches the document; higher levels analyze the content more intensively.

The fourth pipeline creates a new PDF export with cleaned metadata in the background before visual analysis. Use it when a PDF cannot be processed reliably with the first three pipelines.

Using a VLM Pipeline#

The VLM pipelines use AI vision models to understand and extract content from documents. They are particularly useful for:

• Scanned or photographed documents with poor quality.
• Documents with complex visual layouts.
• Handwritten notes or annotations.
• Problematic PDFs that DOCLING or standard visual analysis cannot process reliably.

Retry Embeddings#

If content extraction succeeded but vectorization (embedding creation) failed for a file, you can retry the embedding step without reprocessing the entire file. This is a lightweight operation.