Confluence
Connect Confluence directly in MeinGPT — via API token, with space selection in the UI
Set up in MeinGPT
Open Settings → Data sources, click Add data source, and choose Confluence. First, enter the credentials:
| Field | Required | Description |
|---|---|---|
| Confluence URL | ✅ | e.g. https://company.atlassian.net |
| Username / Email | ✅ | The account the API token below was created with |
| API Token | ✅ | Create one at id.atlassian.com |
Click Fetch Spaces — MeinGPT uses these credentials to list the available Confluence spaces, split into Spaces and Personal Spaces and searchable. Pick the space MeinGPT should make searchable; the source is created directly with that selection.
Confluence runs as a Cloud data source; it needs no Outpost of your own, and this source type is not available for on-premise Outpost deployments.
Synchronization
- MeinGPT reads the page content of the selected space.
- A metadata-based delta sync updates changed pages and removes deleted content.
- Later runs are incremental by default.
Attachments are not ingested
Files attached to a Confluence page are not downloaded by the sync. Where an attachment is linked or embedded in the page text, a placeholder carrying the file name takes its place — so the name is findable, the file's content is not. Attachments that only sit in the page's attachment list leave nothing at all. Typical symptom: the assistant knows the name of a price list but cannot answer anything from it.
Two ways around it if your knowledge sits in attachments:
- As a page instead of an attachment. In Confluence, Import → Word document turns each
.docxinto its own page, whose content is indexed normally. - A file-based source for the files. SharePoint, Google Drive, Amazon S3 and WebDAV read PDF and Office files in full.
The same goes for images: one embedded as an attachment reaches the index only as a placeholder, one embedded by URL leaves nothing at all — see Images in a data pool.