Skip to main content

Configuring a connector

This page covers the general flow for adding a source. Specific fields vary per connector, but every connector follows the same lifecycle: connect → grant → sync → browse.

1. Add the connection​

In the admin UI, open Sources and choose + Add source, then pick the connector type (Confluence, Jira, Git, SQL, Salesforce, …). Provide:

  • the endpoint / instance details for the source system,
  • credentials or a secret reference (secrets are stored write-only and never returned on read), and
  • any scope the connector supports (a Confluence space, a Jira project, a Git repository, a folder).

2. Approve delegated credentials (if applicable)​

Some sources use delegated, per-user credentials. These are default-denied: a pending grant lands in the admin approval queue, and credentials are only released after an admin approves. This keeps user-scoped access auditable and under explicit control.

3. Run the initial sync​

Trigger the first sync. Premora will:

  1. pull content from the source,
  2. capture entitlement snapshots (the source ACLs),
  3. normalize content to canonical form and deduplicate by content hash, and
  4. materialize it into wiki pages, chunks, and the knowledge graph.

The source tree updates live as the sync populates — you can expand connector nodes into a navigable folder/file tree and watch it fill in while syncing.

4. Browse and verify​

When the sync completes:

  • browse the materialized pages in the source tree,
  • run a search and confirm citations and lineage are present, and
  • verify ACL behavior — a user should only see what they are entitled to upstream.

Read-only guarantees​

All source gateways enforce hard read-only egress. Structured-data connectors additionally run only safe, read-only queries against read replicas / reporting schemas. Premora reads from your systems; it never writes back.

Keeping sources fresh​

Connectors re-sync on a schedule (and some support auto-ingest). Re-syncs reconcile changes and deletions: removed source content is reconciled out of the materialized wiki, and freshness is surfaced alongside results so users can judge how current an answer is.