Custom DatabasesDataset Manager (Beta)
Overview
Overview of Dataset Manager (beta) for structured data integration
What Dataset Manager (beta) is
Dataset Manager is the structured-data integration layer in meinGPT. Instead of vectorizing unstructured documents, it operates on tabular data in PostgreSQL datasets.
Core model:
- each dataset is isolated (dedicated database + credentials)
- data is exposed as queryable tables
- assistants access data through tool calls for query, search, and stored queries
Typical use cases
- make CRM/ERP exports usable for assistants quickly
- answer reporting and analytics questions via SQL or natural language
- provide reusable business queries as stable tools for agents
What it currently supports
- file imports: CSV, JSON, SQLite, XML
- schema and table analysis (columns, types, sizes, FK/indexes)
- read-only query execution with query history
- optional per-dataset/table search indexing (Meilisearch)
- stored queries as controlled, reusable tool functions