DATTABI — Connections and Auto-Catalog

In DATTA, connecting a source already means cataloging it. Every external source consumed by DATTABI (Neo4j, OpenSearch, relational databases via JDBC, Kafka and others) goes through the platform's central connection registry (Project Guidelines §7) and, the moment the connection is born, automatic cataloging — ACOC, Auto-Catalog On Connect — discovers and publishes the datasets to the catalog. They show up in DATTABI ready to use, in real time, with no manual mapping step. This guide explains the flow from the DATTABI point of view.

Where to register a connection

All paths lead to the same central registry:

1. Sistema›Conexões — the dedicated administration page.
2. DATTABI — Preparar Dados — the Adicionar conexão button in the "Preparar Dados" (data prep) panel of the dashboard editor.
3. Extract Pipeline, Extract Quick and the Data Catalog — the same shared entry points.

Registration is idempotent: the fingerprint sha256(type + host + port + database + user) guarantees that reusing the same configuration returns the same connection, never a duplicate.

Automatic cataloging on connect (ACOC)

As soon as the connection is successfully created, the platform triggers a background scan that:

1. Introspects the source's schemas, tables, labels and indexes.
2. Collects quality metrics — freshness, completeness, distinct-value counts and personal data (PII) detection.
3. Keeps a 100-row sample for preview.
4. Publishes every discovered dataset to the Knowledge Catalog.
5. Emits the audit events CONNECTION.AUTO_CATALOG_STARTED | COMPLETED | FAILED.

There is no opt-out for regular users. An administrator can disable cataloging for an individual connection from the interface itself, with a recorded justification.

How DATTABI consumes the sources

• The list of available datasets is filtered by workspace and by your permissions — you only see what you are allowed to see.
• Each dataset carries its base DATTAX script, which is the initial query (FROM GRAPH, FROM JDBC, FROM INDEX, etc.). When you build a visual, the platform wraps that query with the bindings you defined (X and Y axes, aggregation, filters and drill) and generates the final DATTAX.
• Render results are cached for 5 minutes, per chart and filter combination — reopening the same dashboard is instant.

The API paths for listing the workspace's datasets, registering connections and publishing datasets to the catalog are in the API reference.

Credentials

Passwords never travel or rest in plain text. Every connection registration with a password goes through the platform's credential vault, which creates a per-user binding — and no platform response ever returns the password.

Scope and visibility

• Per workspace: datasets and connections belong to a workspace; whoever has the DATTABI_VIEW permission only sees the workspaces they are a member of.
• Global administrator: whoever has PLATFORM_ADMIN sees everything, for governance and support.

Anti-patterns (rejected in review)

• A component that opens a connection directly, without going through the central registry.
• Hardcoding a Neo4j or OpenSearch address in a specific component's configuration.
• Creating a connection without triggering the cataloging scan (ACOC is mandatory).
• Storing a password in a configuration file.