Data connections overview

A data connection links your data warehouse to Neuralift so it can read the input data for segmentation runs. You set connections up once, per organization, under Settings → Data Connections; after that, every use case draws on the source tables you copy in through them.

Connections are managed by your organization’s admins; see Roles & permissions for the full breakdown of who can do what.

How data flows in

Your warehouse  →  Data connection  →  Copied source tables  →  Data preparation  →  Segmentation run

You connect a warehouse, browse the tables it exposes, and copy the ones you want into Neuralift as source tables. Data preparation then profiles a source table into the dataset a run works from.

Supported sources

Source	How it connects	What you provide	Availability
Databricks	Delta Share: you share tables to Neuralift; no credentials leave your side	A Delta Share created in your workspace, plus its provider and share names	Generally available
Snowflake	Direct connection with a database user	Account identifier, warehouse, database, username, and password	Generally available
BigQuery	Direct connection with a service account	GCP project ID and a service account key (JSON)	Preview; see Release stages
Redshift	Direct connection with a database user	Host, port, database, username, and password	Preview

An organization can hold more than one connection, and each connection has a name you choose (lowercase letters, digits, and dashes; unique within the organization).

Each source page also covers receiving results back into that warehouse — what to provision on your side so published runs land as queryable tables or files in your environment. The delivery workflow itself is on Delivering results to your warehouse.

Note. Which warehouse types appear in the Add data connection dialog depends on what’s enabled for your organization. If a source you need isn’t listed, contact support@neuralift.ai.

Connection statuses

Each connection card in Settings → Data Connections shows a status badge:

Status	What it means
Not configured	The connection has been created but setup hasn’t completed yet.
Setting up…	Shown as a brief notice while Neuralift provisions a new connection. The dialog closes, a Setting up … connection… notice appears, and after a moment the card appears with an Active or Error badge. No action needed.
Active	The connection is live. You can Browse tables, Test it, and copy source tables into Neuralift.
Error	Setup or a later test failed. The card shows a plain-language explanation of what went wrong.

On an Active connection, the Test button re-checks connectivity on demand, and the card shows when the connection was last tested.

What Neuralift stores

Neuralift’s own database stores zero warehouse secrets. What it keeps is the connection’s display metadata: its name, warehouse type, and non-secret details such as a Snowflake account identifier or a Redshift host. Credentials you enter (usernames, passwords, service account keys) pass straight through to Neuralift’s governed compute layer, where they are encrypted — and re-saving a connection with new credentials rotates them in place. Delivery back to your warehouse involves no shared secrets at all: it authenticates with identities — metastore IDs, one-time activation tokens, cross-account IAM roles, and per-organization service accounts. Databricks Delta Share connections involve no credentials in either direction: you share data to Neuralift from your side, and results are shared back the same way. The full picture is in Data handling & security.

Removing a connection disconnects it from Neuralift’s environment; the confirmation dialog spells out exactly what will be torn down.

Next steps

• Connect your warehouse: Databricks, Snowflake, BigQuery, or Redshift.
• Then copy in the source tables you want to segment on.
• Curious what happens to the data next? See the data preparation overview.