Load data into DuckDB first, then let the Database Agent query it.
The Database Agent imports project data into the default DuckDB, inspects the schema, runs SQL, and organizes results into reusable tables. After a project opens, the database is at .agents/db/qiyan.duckdb. Supported formats include CSV, TSV, TXT, XLSX, DTA, SAS, SPSS, JSON, JSONL, NDJSON, and Parquet.
Applies to QiyanAgent V3.5. Last updated: September 14, 2026.
This is a QiyanAgent V3.5 interface screenshot. The callouts mark data import, database refresh, task entry, and task submission.
Before You Start
- Confirm the data files to import are in the project folder.
- Sample the file first to inspect field names, delimiters, encoding, and obvious bad rows.
- Prepare clear table names so later analysis, visualization, or writing tasks can reuse them.
Steps
- Click Import Data at the bottom of the left database view, choose a data file in the project folder, and let the import run as a background task.
- After import, expand the schema to review table names, field types, row counts, and fields. Any bad row, parsing error, or type-conversion error rolls back the entire import.
- Review Database Agent Skills under Settings → Skills. Database tools are supplied automatically for the active profile and are no longer managed manually in a right-side Tools area.
- Enter a database prompt with the file to import, target table name, query definition, or output format.
- Click the send button and let the Database Agent run SQL, aggregate, filter, join, or export results.
Acceptance Checks
Query results should include the SQL definition, result table, row-count notes, a reusable table name, and data-quality warnings. Export currently supports CSV and XLSX through a background task to a path chosen by the user. When handing results to another Agent, state the target table, field boundaries, and filters.