Connect Confluence, SharePoint & Slack

This guide wires an external source into the knowledge base as a content connector: create the source with credentials, run the first full backfill, and let incremental sync keep it fresh. Everything it ingests becomes documents you can search and cite exactly like a posted one.

Prerequisites

• A key with kb:write (create/sync a source) and data:read (list sources), exported as $ELI_KEY. See Getting started.
• Owner/admin on the workspace — creating connectors and storing credentials is owner/admin-gated.
• The per-source credential below, from the upstream system's admin.

Per-source setup

Confluence

Create an API token for a read-only service account, and note the space keys to sync and your base URL. The connector polls by lastModified via CQL and paginates with the v2 cursor, so incremental syncs only pull pages edited since the last watermark.

{
  "type": "confluence",
  "config": {
    "baseUrl": "https://acme.atlassian.net/wiki",
    "spaceKeys": ["ENG", "OPS"]
  },
  "credential": {
    "email": "svc-eli@acme.com",
    "apiToken": "<atlassian-api-token>"
  }
}

SharePoint

Register an Entra ID (Azure AD) app with the application Microsoft Graph permission Sites.Selected and grant tenant admin consent. That permission starts with access to no sites: a SharePoint/Graph administrator must separately grant the connector app read on every site id you configure. The connector uses the Graph driveItem delta query and stores the opaque deltaLink in sync_state, so each incremental call returns exactly what changed — additions, edits, and deletes.

POST https://graph.microsoft.com/v1.0/sites/<site-id>/permissions
Authorization: Bearer <site-grant-admin-token>
Content-Type: application/json

{
  "roles": ["read"],
  "grantedToIdentities": [{
    "application": {
      "id": "<app-client-id>",
      "displayName": "eli.ai connector"
    }
  }]
}

{
  "type": "sharepoint",
  "config": {
    "siteIds": ["acme.sharepoint.com,<site-guid>,<web-guid>"]
  },
  "credential": {
    "tenantId": "<entra-tenant-id>",
    "clientId": "<app-client-id>",
    "clientSecret": "<client-secret>"
  }
}

Slack — a customer-owned internal app

1. Create the app from the manifest

   In api.slack.com/apps, create a new app from the provided manifest. It requests read-only bot scopes: channels:history, channels:read, groups:history, users:read (to resolve author names).

2. Install and copy the bot token

   Install the app to the workspace and copy the Bot User OAuth Token (xoxb-…). Invite the bot to the channels you want synced — it can only read channels it is a member of.

3. List the channel ids

   Note the channel ids to sync. The connector keeps one cursor per channel in sync_state and walks history with cursor pagination and backoff that respects the tier; Slack mrkdwn is transformed to markdown locally.

{
  "type": "slack",
  "config": {
    "teamId": "T0XXXXXXX",
    "channelIds": ["C0AAAAAAA", "C0BBBBBBB"]
  },
  "credential": { "botToken": "xoxb-<bot-token>" }
}

1 · Create the source

POST the config and credential. The secret is encrypted at write (AES-256-GCM, enc:v1); the response masks it and every later read returns only hasCredential.

curl -s -X POST https://your-host/api/v1/connectors \
  -H "Authorization: Bearer $ELI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "confluence",
    "config": { "baseUrl": "https://acme.atlassian.net/wiki", "spaceKeys": ["ENG"] },
    "credential": { "email": "svc-eli@acme.com", "apiToken": "<atlassian-api-token>" }
  }'

{
  "id": "01KZCONNSRC00000000000000",
  "type": "confluence",
  "status": "active",
  "config": { "baseUrl": "https://acme.atlassian.net/wiki", "spaceKeys": ["ENG"] },
  "hasCredential": true,
  "lastSyncedAt": null,
  "createdAt": "2026-07-21T09:14:03.412Z"
}

2 · Kick off the first full sync

A fullsync backfills everything. It persists every document in a page before advancing that page's checkpoint, so interruption cannot acknowledge content that was only queued. A PostgreSQL advisory lock keyed by workspace and source serializes full, incremental, prune, and permissions work even when different queues or worker replicas receive overlapping jobs.

curl -s -X POST https://your-host/api/v1/connectors/01KZCONNSRC00000000000000/sync \
  -H "Authorization: Bearer $ELI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"kind":"full"}'

{ "enqueued": true, "sourceId": "01KZCONNSRC00000000000000", "kind": "full" }

Each document is upserted by (source_id, external_id), so a re-sync updates rather than duplicates, and content-hash detection skips re-embedding anything unchanged. Every connector change, reactivation, and prune delete remains in document revision history. Watch progress in the source's sync_runs history (docs seen / added / updated / successfully tombstoned) on the Connectors page.

3 · Keep it current

After the backfill, incremental syncs do the steady-state work — Confluence by CQL watermark, SharePoint by Graph delta, Slack by per-channel cursor — and a prune pass tombstones anything deleted upstream. Trigger an incremental manually the same way ({"kind":"incremental"}) or let the schedule run it. A failed document leaves the page checkpoint unchanged; retry replays prior successful upserts idempotently. Prune first completes the authoritative remote id listing, then counts a tombstone only when its soft delete succeeds.

curl -s https://your-host/api/v1/connectors \
  -H "Authorization: Bearer $ELI_KEY"

{
  "connectors": [
    {
      "id": "01KZCONNSRC00000000000000",
      "type": "confluence",
      "status": "active",
      "hasCredential": true,
      "lastSyncedAt": "2026-07-21T09:22:41.008Z"
    }
  ]
}