Skip to main content

Confluence to Pinecone

This workflow connects Confluence Cloud, prepares selected pages with your embedding model, stores the vectors in Pinecone, and verifies retrieval in the Playground.

Expected result​

The selected Confluence pages reach Synced status in one Vector Data Loader project, and a specific phrase from a known page returns that page from the project's Pinecone index.

Before you start​

You need:

  • A Vector Data Loader account with permission to manage sources and project settings
  • Access to the Confluence Cloud spaces and pages you intend to sync
  • An embedding provider API key and a Pinecone API key
  • A Pinecone dense-vector index whose dimension matches the selected embedding model

Do not paste an Atlassian access token into Vector Data Loader. The built-in Confluence connection uses Atlassian OAuth 2.0 authorization.

1. Choose the project and embedding model​

  1. Select the project that should own the documents.
  2. Open Settings → Embedding Model.
  3. Select the provider and model.
  4. Save the provider API key and test the connection.
  5. Record the model's vector dimension.

Checkpoint: The embedding connection succeeds and the configured vector dimension is known.

2. Configure Pinecone​

Vector Data Loader creates embeddings before sending vectors to Pinecone. Use a Pinecone dense-vector index with the same dimension as the configured embedding model. Pinecone also recommends choosing a similarity metric appropriate for the embedding model.

  1. Create or choose a compatible dense-vector index in Pinecone.
  2. In Vector Data Loader, open Settings → Vector Store.
  3. Select Pinecone.
  4. Enter the index name and, if needed, an optional namespace.
  5. Save the Pinecone API key.
  6. Choose Test Connection.

Checkpoint: The connection test succeeds and the displayed index dimension matches the embedding model.

Dimension mismatch

Do not begin a large sync when the Pinecone index and embedding model use different dimensions. Pinecone will reject vectors that do not match the index schema.

3. Connect Confluence Cloud​

  1. Open Sources.
  2. Choose Add Source, then Confluence.
  3. Complete the Atlassian consent flow.
  4. Choose an Atlassian site where your account can read the required content.

Atlassian scopes limit what the integration may request. Your Confluence permissions also remain in effect, so Vector Data Loader cannot expose a space or page that the connecting user cannot read.

Checkpoint: The source appears as connected without asking you to supply an Atlassian token.

4. Browse and sync pages​

  1. Choose Browse on the Confluence source.
  2. Expand a space to load its root pages and folders.
  3. Expand only the branches you need. Content loads as you open each level.
  4. Select current pages to sync.
  5. Start the sync and confirm the target project.

Archived, trashed, and draft pages do not appear in the browser. A parent selection covers the direct pages currently loaded; open nested branches when you need their descendants.

Checkpoint: The selected pages appear in Documents for the intended project.

5. Verify retrieval​

  1. Wait for the required documents to show Synced status.
  2. Open Playground → Search.
  3. Search for a distinctive phrase or concept from one synced page.
  4. Inspect the returned title, source link, excerpt, and relevance score.

Checkpoint: The known Confluence page appears for the test query. If it does not, verify the active project, embedding configuration, Pinecone connection, and document status before adding more content.

Primary provider documentation​

See also Confluence Integration, Vector Store Configuration, and Search.