Formats & Access
Supported file formats, emails and attachments, and access control for data pools
Which file formats a data pool makes searchable, and who can access a data pool. Setup and configuration of a data source: Data Sources & RAG. What you can do to the files beforehand: Preparing Your Data.
File formats
Well supported are all formats that primarily consist of text:
- Office documents: DOCX, PPTX, ODT (OpenDocument Text)
- Spreadsheets: XLSX and XLS (with caveats, see below), and CSV
- Email files: EML and MSG (see "Emails and attachments" below)
- TXT, Markdown, RTF, EPUB, HTML
- Image files: PNG, JPG, GIF and WEBP (a vision model describes them, see Images in a data pool)
What gets skipped during sync
- File extensions not listed here. Only the formats above are made searchable. Files with other extensions (for example
.doc,.ppt, or.json) are not added to the data pool during sync - they appear neither in the file list, nor as an error in the Sync Logs, nor in the count of skipped files. If an expected file is missing, check its extension against the list above first. Convert old Office formats to.docxor.pptxbeforehand, see Preparing Your Data. - Some image files. A cloud data source reads PNG, JPG, GIF and WEBP; TIFF, BMP, SVG and other image formats are skipped. Images inside documents are read - how that works, how to lay images out for it, and how an image becomes visible in chat is covered in Images in a data pool.
- Encrypted or rights-managed PDFs. If a PDF file is password-protected or carries a rights-management (DRM) profile, MeinGPT can't read its content or make it searchable. Remove the password protection, or export the document without the DRM restriction, before adding it to the data pool.
- Files with an encrypting sensitivity label (Microsoft Purview / MIP). If a sensitivity label applies encryption to a file, it stays encrypted in SharePoint. MeinGPT cannot read its content or make it searchable, even if the connected account can open the file in SharePoint itself. There is no setting in MeinGPT that changes this. Such files usually show up in the data source's Files tab as Password-protected. Put content that MeinGPT should search into a separate library or folder whose files carry a label without encryption, and point the data source at it. Agree this with your IT or data protection team.
- Excel tables are a special case. A table is not split into passages like running text. What is captured is its structure (sheet names, title rows, column headers, sample rows) and the first data rows, each with its column headers: at most 2,000 rows per sheet, 5,000 rows per file, and 64 columns per row, fewer for wide tables or long cells. Search does not find rows beyond that. For calculations, analyses, and visualizations from Excel files, use the Code Sandbox instead. It processes the original file directly.
Limits per file
The 50 MB per file from chat and assistant upload is not the yardstick for data sources. Files in a cloud data source go through their own processing with their own limits, and a file can be rejected well below 50 MB.
- Scanned PDFs: at most 50 pages. MeinGPT reads PDFs without a text layer with text recognition. A scanned PDF with more than 50 pages is not processed and shows up in the data source's Files tab as Too large, even though file size is not the problem. Add a text layer to the PDF with an OCR tool before uploading. Splitting into parts of at most 50 pages alone is not reliable for some scans. This page limit does not apply to PDFs that have a text layer.
- Other processing limits. Very large workbooks and documents with an extreme amount of text are rejected as well and listed as Too large.
- Very large files: partially indexed. For files of roughly several hundred pages of running text, meaning search covers only a selection across the whole file. The file then shows the note Partially indexed. Keyword search still covers the whole file, so a passage can be found by its words. Only for extremely large files does keyword search also cover just the beginning. The file is not rejected and can still be read in full. Split collective documents by chapter, see Preparing Your Data.
What the labels in the Files tab mean: Sync Status & Troubleshooting.
OneNote workaround. OneNote files are currently not indexed directly because the format is proprietary. Workaround: export OneNote content regularly via Make or n8n, as PDF or text. The exported files can then be connected like any other source.
Attention
OneNote notebooks don't show up in the sync status - not even as "skipped". Like every unlisted file type, OneNote notebooks and sections appear neither in the file list, nor in the Sync Logs' error messages, nor in the count of skipped files (see Indexing progress). If the connected SharePoint source also contains other supported files, those still index normally - so the sync status can show something like "1 indexed - No errors" even though the OneNote notebook itself never made it into the index. A clean sync status is therefore not proof that OneNote content is searchable. Check the data source's file list directly for the files you expect, or use the export workaround described above.
Emails and attachments
You can place email files in EML (open standard) and MSG (Outlook) format directly into a data pool. The subject, sender, recipients, and message body are extracted and made searchable. Both formats are supported equally - you do not need to convert MSG to EML beforehand.
Attention
Attachments inside email files are a special case. For smaller emails, the content of attachments (e.g. a PDF) is read and indexed. For larger emails, attachments are reduced to their filename for performance reasons - their content is then not searchable. If you want a PDF attachment indexed as reliably as a PDF that sits directly in the data pool, also add the attachment as its own file to the data pool.
Access control
A newly created data pool is shared with nobody except admins (in a decentral workspace it starts out shared with its creator only) - nobody else has access automatically, not even through a linked assistant. To give others access, share the pool explicitly: with individual users, a team, or workspace-wide for everyone - currently always as view-only access ("Kann ansehen" in the sharing dialog). You can additionally restrict access to specific teams: create teams in the admin interface and assign them to specific data sources. This lets you control which user groups see which data.
Details on creating and managing teams: Team management.
Review and trace sharing
Every change to who a data source is shared with (a person, a team, or everyone in the workspace) appears in the audit log under Settings → Audit log, with who, when and the new value. That shows you who set a source to "everyone". You can export the log (see Export the audit log) for analysis.
Under Settings → Data sources, admins see the Review sharing notice when data sources are shared with no one or only with individual people. From there you can share each source with a team, share it with everyone after a confirmation, or mark its sharing as reviewed with Leave sharing as is.
Attention
Sharing an assistant does not override a data pool's own access restriction. If a data pool is not shared with everyone - for example only with specific users, a team, or admins - that restriction applies regardless of who the associated assistant is shared with. Someone who can open the assistant but has no access of their own to the data pool gets nothing from it. MeinGPT says so openly: below the answer it shows e.g. "1 source of this assistant is not shared with you", and the assistant names only the source - never anything from its content. Whoever shares an assistant more widely than one of its sources sees the note "Colleagues without access to … find nothing in it" under "Knowledge" in the assistant editor.
Who may create data sources?
Under Settings → Data sources, admins decide who creates data sources:
- Central (default): Only admins create data sources and share them with the team. "Settings → Data sources" is then an admin-only area.
- Decentral: Every member can also create their own data sources. Members see only their own data sources under Settings → Data sources. A new data source starts out visible only to its creator; they can share it as usual and connect it to their own assistants.
Admins still see and manage every data source. Switching does not change existing data sources: if the workspace goes back to "Central", members' data sources keep their shares and are managed by admins from then on. Every switch is recorded in the audit log.
The choice only appears in workspaces for which MeinGPT has enabled the decentral mode. Without it, the workspace is central.
You can only add data sources to an assistant that you have access to yourself.
Independently of that, members also see the data sources they are responsible for under Settings → Data sources.
Automatic data source search
Under Settings → Data sources, admins decide whether a normal chat without an assistant and without a chosen source automatically searches every data source the person may use:
- Always off (default): The chat only searches a source chosen via Search a data source, or an assistant's sources.
- Members decide: Each member turns on Search data sources automatically themselves, in the chat's "+" menu or under Preferences. The switch starts off and applies to this workspace only.
- Always on: Without a choice, the chat can search every data source the person may use. The model decides per question whether to search. This uses more credits, and answers draw on company data more often.
With Always off and Always on, the switch is greyed out for members and shows "Set by your admin". A source chosen in the chat and an assistant's sources are searched either way. Only the admin setting is recorded in the audit log, not a member's own switch.
Responsible people for a data source
Admins do not have to look after every data source themselves. On a data source's page, an admin assigns individual people or teams under Settings → Responsible people who may manage exactly that source.
Responsible people can, for their source:
- change its name and purpose, and for Outpost sources the passages per answer
- trigger Sync now and see its status, files and activity
- use the test search, including "as" another member — results with document text only when the source is also shared with them
- share it with people and teams, or remove those shares
Responsible people cannot delete a data source or change its connection (a stored account would otherwise read another place) — that stays with admins and, in a decentral workspace, with the person who created the source. Only admins create sources in a central workspace, share a source with everyone in the workspace, lift a stored release restriction, assign further responsible people, and change workspace settings such as permissions from the source.
Being responsible grants no search access — who finds the content in a chat is still decided by sharing. What responsible people change or share is recorded in the audit log just as for admins. The default "Everyone" team cannot be responsible; that is what admin access is for.