Configure a connection first, then let the Remote Database Agent query it.
The Remote Database Agent securely connects to MySQL, Doris, or Oracle in the project context, inspects schemas, generates SQL, filters and aggregates data, and passes results to project-database, analysis, or visualization workflows.
Applies to QiyanAgent V3.5. Last updated: September 14, 2026.
This is a QiyanAgent V3.5 interface screenshot. The callouts mark connection setup, connection selection, task entry, and task submission.
Before You Start
- Prepare the database address, port, database name, username, and password under Settings → Database Connections.
- Confirm that the account permissions match the task, prefer a read-only account, and never put the password in project files or prompts.
- State the target tables, field meanings, time range, filters, and expected output so the Agent does not scan unrelated data.
- Oracle users must configure Thick mode for the environment. For Oracle 11g compatibility issues, check the client and driver configuration first.
Steps
- Open Settings → Database Connections, create a MySQL, Doris, or Oracle connection, complete the connection test, and save it.
- Choose read-only or read-write access. Read-only is suitable for querying, sampling, and analysis; use read-write only when writing is explicitly needed and a backup exists.
- Switch to the Remote Database workspace on the left and choose a saved connection from Select Remote Database Connection.
- Click New SQL or describe the task to the Remote Database Agent at the bottom. Ask it to inspect the schema and a small sample before generating the query.
- Before running read-write SQL, the software saves it to the current project's
sql/remote_db_history/; execution stops if saving fails. - When reusing results, pass filtered, aggregated, and redacted data to the project database or export it to
data_exports/instead of copying the whole database.
Acceptance Checks
Results should state the connection, table and field scope, SQL definition, filters, returned row count, and data-quality warnings. Read-write tasks should have a pre-execution statement in the project SQL history. Redact personal or sensitive fields and review permissions.