Bring your warehouses, databases, cloud storage, and the apps your team already uses into Summation.
Connectors are how your data reaches Summation. Connect a source once and its data stays current, so everything you build on top of it (chats, artifacts, scheduled workflows) reads live from the system your team already trusts.Connectors come in two kinds, on two tabs of the same page:
Data: warehouses, databases, object storage, and repositories. Pick which tables or files to bring in as datasets, and those datasets become available to Addison, artifacts, and your tables and data catalog.
Apps: third-party tools like Shopify, PostHog, or Tableau. Connecting one lets Addison work with that app directly in a chat.
The Connectors page: your data connections and their status
Click New connection (top-right of the Connectors page) to start a three-step wizard.
1
Choose a data source
Pick a connector type. Snowflake, Postgres, and BigQuery are surfaced under Popular; everything else is under Other, and app integrations appear under Apps.
Step 1: choose a data source
2
Enter connection details
Give the connection a Name and an optional Description. The name is the identifier used in SQL and the API, so it must start with a letter and contain only letters, numbers, hyphens, and underscores. Then fill in the connector’s own fields; View the setup guide links to that connector’s page below.You can also turn on Scheduled snapshots here. Datasets inherit the connection’s schedule unless they override or disable it.Click Test connection before continuing. A pass means Summation could authenticate and reach your data; a failure shows the message the source returned.
Step 2: connection details and credentials
3
Add datasets
Pick the tables, files, or repositories to expose. The browser is tailored to the connector: a database, schema, and table tree for warehouses, a file browser for object storage, a repo picker for GitHub. It has a tree on the left, search on the right, and a running count of what you’ve selected.Each selection becomes a dataset. Dataset names must be unique across your tenant, and the wizard suggests a unique name when there’s a conflict.
Click any row on the Data tab to open its detail page.
Configuration & stored secrets: the non-secret fields (project, host, account) and which secrets are stored. Secret values are never shown after save; they render as ****.
Test connection: re-check credentials and reachability at any time.
Edit: update credentials when they rotate, or change configuration such as the warehouse, role, or region.
Disconnect: disconnect every dataset in this connection at once, leaving the definitions in place. Other connections aren’t affected, and you can reconnect later.
Delete: permanently remove the connection, its credentials, and all of its datasets.
A connection's detail page
Connection statuses. The colored dot on each connection shows its current state.
Status
Color
What it means
Active
Green
The connection is healthy and its datasets are queryable
Processing
Yellow
Datasets are deploying or syncing in the background
Error
Red
The last operation failed. Open the connection to see what went wrong
Disconnected
Grey
The connection is paused. Datasets stay defined but no queries run until you reconnect
A connection’s Datasets tab lists everything it brings in: the dataset name, its source reference, when it refreshes, how the last refresh went, and the sync Strategy. Use + Add dataset to bring in more, the per-row refresh button to sync one now, and the Refresh history tab to see past runs across the connection.Sync strategies. How each refresh updates the stored copy.
Strategy
What it does
Full refresh
Replaces the dataset with a complete re-read of the source
Incremental (append)
Reads only new rows since the last run and appends them
Incremental (merge)
Reads only what changed and merges it into existing rows
Incremental (window)
Re-reads a moving time window and replaces just that slice
Incremental strategies need a column to track progress by, typically an updated-at timestamp or an increasing id.
The Apps tab lists app integrations grouped by category: AI, Analytics, Commerce, and more. Click + on an app to review what it does, then Connect. Most apps hand off to that provider’s sign-in to authorize access. Once connected, Addison can work with the app directly in a chat.
Available apps, by category
Review an app before connecting
Connected apps move to a Connected section at the top of the tab, where you can view the connector’s details or Disconnect it. If the app you need isn’t listed, Request an app tells us what to add, and a custom MCP server URL connects one we don’t list.
Third-party apps aren’t built or maintained by Summation, so review the permissions before connecting. Summation doesn’t train models on your connector content, and you can disconnect at any time.
Apps
Available apps by category, connecting one, and adding a custom app
Test connection fails immediately. Double-check the host, the port, and that your network or firewall allows traffic from Summation’s egress IPs.
Test connection succeeds but datasets fail to load. This is usually a permissions problem on the source side. Check that the user or role you provided has read access to the schemas and tables you want.
Secrets seem to disappear after editing. Secret fields are write-only. They render blank after save, but the stored value is still there until you overwrite it.
Dataset name conflicts. Names must be unique across the tenant. Rename in the wizard before deploying.
A dataset shows a failed refresh. Open the connection’s Refresh history for the run’s error, then re-run that dataset with its row refresh button.
If a problem persists, copy the error from the Test connection result and contact support.
Assistant
Responses are generated using AI and may contain mistakes.
