Building a Data Assistant Application with DB‑GPT V0.6.0

The article introduces DB‑GPT V0.6.0 and explains how to build a data‑assistant application that combines intelligent question answering, database‑driven data dialogue, and internet search capabilities through three core abilities: document‑based QA, database query dialogue, and tool‑based web search.

Prerequisites : Deploy the DB‑GPT project, configure the environment, and refer to the provided deployment documentation (https://www.yuque.com/eosphoros/dbgpt-docs/qno7x8hmgyulbg80).

Sub‑agent construction : Create three sub‑agents (Intent Recognition Expert, CodeEnginner, Reporter, DataScientist, Summarizer, ToolExpert, …) and integrate them via AppLink. The example focuses on a RAG‑based QA assistant that uses the Summarizer agent together with a domain knowledge base built from the OceanBase official PDF.

Knowledge‑base creation : Upload the PDF document to the DB‑GPT knowledge‑base UI. Select the Embedding storage type. Complete the upload and indexing steps (see screenshots in the original guide).

QA application creation : In the Application Management console, create a new app, choose the Summarizer agent, optionally edit the prompt, set model priority, attach the previously created knowledge base as a resource, and add a recommended question.

Data dialogue assistant : Similar steps are followed to create a DataScientist‑based app that queries a configured database and can render charts.

Search assistant (AWEL workflow) : List remote AWEL examples with dbgpt app list-remote . Install the web‑info‑search workflow using dbgpt app install awel-flow-web-info-search . Verify installation in the AWEL UI and configure the workflow (Agent Resource, ToolExpert, Summarizer).

Intent‑recognition knowledge base : Define intents (e.g., DB‑QA, Data Dialogue, Weather Assistant) with fields Intent, App Code, Describe, and Slots, using ‘#’ as a delimiter. Upload the intent file to a new knowledge base.

Unified intelligent application : Combine all sub‑agents into a single entry point by leveraging intent‑recognition and AppLink routing. Install the provided intent‑recognition AWEL workflow ( dbgpt app install db-expert-assisant ), copy and customize it, and link the Intent Recognition Expert, AppLauncher, and Summarizer agents.

Finally, create the unified app in task‑flow mode, select the customized workflow, add recommended questions, and start conversations to test the integrated data‑assistant functionality.