External knowledge sources let Copilot read context from outside SqlDBM; specifically from Confluence pages and Slack channels that contain your team’s data documentation. This article covers how to connect an external source, what Copilot does with external context, and the recommended interim pattern for teams that need this today.
Feature availability: External knowledge source integration is planned for a future release. Until it ships, the best interim pattern is to capture governance context in SqlDBM data governance pages, which Copilot reads natively. See “Data Governance: Pages” for details on that workflow.
What external knowledge sources do
Many teams document data context in places other than SqlDBM: Confluence for long-form definitions, Slack for point-in-time conversations and decisions, Jira for change tickets. When a user asks Copilot a question like “how is customer lifetime value calculated,” the answer is often not in the schema, it’s in a Confluence page describing the business rules.
External knowledge sources let Copilot read those sources as part of its context. When a user asks a question, Copilot can pull relevant Confluence pages or Slack threads alongside the schema metadata and generate a response grounded in the full picture.
How the integration works (conceptually)
The integration is designed around three principles:
- Permissions are respected. Copilot can only read pages and channels the user has access to. A consumer asking about a customer dimension won’t accidentally see restricted finance documentation.
- Context is scoped, not exhaustive. Copilot fetches relevant pages per question, not the entire Confluence space. This is how the system manages context windows on large knowledge bases.
- Freshness is automatic. When a Confluence page is updated, the next Copilot question that references it will pull the current version. There’s no caching or manual refresh step.
Connecting Confluence (when available)
The setup flow will follow the standard enterprise integration pattern:
- In SqlDBM, open Account Settings > AI Copilot > External Sources.
- Click Connect Confluence.
- Authorize via OAuth: SqlDBM requests read-only scopes.
- Select which Confluence spaces Copilot should read. Leave off any space that shouldn’t be considered.
- Save. Copilot begins using the Confluence context on the next prompt.
Context window note: AI models have a limit on how much context they can process per call. If you select many large Confluence spaces, the system prioritizes the most semantically relevant pages for each question rather than loading everything. Start with the spaces where your data documentation lives; you can add more over time.
Connecting Slack (when available)
Slack integration works similarly but is more selective by default. Slack contains much more signal than Confluence, and most of it isn’t data documentation. Recommended pattern:
- Connect Slack via OAuth.
- Select specific channels Copilot should read. Typically channels named things like \#data-questions, \#data-governance, or \#analytics-help.
- Do not connect general-purpose channels. Copilot has no way to know that a joke in \#random isn’t data context.
- Review Copilot’s use of Slack context after a few weeks. If the AI is citing Slack in unhelpful ways, narrow the channel list further.
The interim pattern: use SqlDBM data governance pages today
Until external integrations ship, the strongest pattern for getting Copilot to read your data documentation is to put that documentation inside SqlDBM’s data governance pages. Data governance pages:
- Are natively part of Copilot’s context.
- Can be linked to specific tables and columns so context is discoverable.
- Support rich text, headings, and embedded objects.
- Can be imported from Word or exported to PDF.
- Respect the same permissions as the rest of the project.
For teams with existing Confluence documentation, a common onboarding step is to copy the most-referenced pages into SqlDBM data governance pages. It’s not automatic, but it moves the documentation that matters most into a place where Copilot can read it today. When external integrations ship, the pages you’ve already created continue to work. You haven’t wasted the effort.
When to use external sources versus internal pages
- Use SqlDBM data governance pages when: the documentation is specifically about the data model, changes relatively rarely, and benefits from being co-located with the objects it describes.
- Use Confluence integration when: the documentation is general-purpose (not data-model-specific) and is maintained by a broader organization for reasons unrelated to SqlDBM.
- Use Slack integration when: your team makes data decisions in specific designated channels and wants Copilot to surface recent consensus. Be selective about which channels you connect.
Most teams end up using a combination. The most business-critical documentation lives in SqlDBM pages (where modelers can maintain it alongside the schema); enterprise-wide context sits in Confluence; real-time conversation lives in Slack.
Related articles
- Getting started with Copilot
- Configuring Copilot settings and pre-prompts
- Data Governance — Pages
- Talk to your model