Chat

DB-GPT provides multiple chat modes through its Web UI — each tailored for a different use case.

Chat modes

Mode	Description	Requirements
Chat Normal	General conversation with the LLM	LLM configured
Chat Data	Natural language queries on SQL databases (Text2SQL)	Database connected
Chat Excel	Upload and query Excel/CSV files	LLM configured
Chat Knowledge	RAG conversations over documents	Knowledge base created

Starting a chat

1. Open the Web UI at http://localhost:5670
2. Click Chat in the sidebar
3. Select a chat mode from the dropdown or create a new conversation
4. Type your message and press Enter

Quick test

Start with Chat Normal to verify your LLM is working correctly, then try other modes.

Chat Normal

The default mode — a direct conversation with the configured LLM.

Features:

• Multi-turn conversation with context retention
• Markdown rendering for formatted responses
• Code syntax highlighting
• Streaming responses

Chat Data (Text2SQL)

Query your connected databases using natural language. DB-GPT converts your question into SQL, executes it, and presents the results.

How to use:

1. First, connect a database in the Database section of the sidebar
2. Start a new Chat Data conversation
3. Select the target database from the dropdown
4. Ask your question in natural language

Example:

User: Show me the top 10 customers by total order amount
DB-GPT: [generates SQL, executes, and displays results in a table]

Supported databases

MySQL, PostgreSQL, SQLite, ClickHouse, DuckDB, MSSQL, Oracle, and more. See Data Sources for the full list.

Chat Excel

Upload an Excel or CSV file and query it with natural language.

How to use:

1. Start a new Chat Excel conversation
2. Upload your file (.xlsx, .xls, or .csv)
3. Ask questions about the data

Example:

User: What is the average sales amount per region?
DB-GPT: [analyzes the file and presents results]

Chat Knowledge

Conversational RAG — ask questions and get answers grounded in your uploaded documents.

How to use:

1. First, create a knowledge base and upload documents (see Knowledge Base)
2. Start a new Chat Knowledge conversation
3. Select the knowledge base from the dropdown
4. Ask your questions

Features:

• Answers cite source documents
• Supports multiple file formats (PDF, Word, Markdown, TXT, etc.)
• Combines vector search with LLM generation

Conversation management

• History — Previous conversations are saved in the sidebar
• Delete — Right-click a conversation to delete it
• Export — Copy conversation content from the chat window

Next steps

Topic	Link
Set up knowledge base for Chat Knowledge	Knowledge Base
Connect databases for Chat Data	Data Sources
Build custom chat workflows	AWEL Flow