Conversational SQL Node

Definition

The Conversational SQL node is used in conversational workflows to perform CRUD operations on a database. It accepts a specified database table and user input, supports intelligent SQL generation from natural language, and passes results to downstream nodes.

---

Configuration

Add a Database

Click the + icon next to the database field, then click New Database. Enter a valid database name and description, then click confirm to enter the database table structure design page.

1. Add table structure — Fill in field names and descriptions, select data types, and mark required fields.

2. Import data — After creating the table structure, switch to the data import tab. Upload an Excel or CSV file, map the table fields, and complete the import.

3. Preview data — After import, switch to the table preview tab. A global search box is available at the top left, and imported data is displayed in a paginated list below.

Add a Database to the Node

In the database dialog, click Add to add the database table to the node's data table list. You can add or remove tables as needed.

Remove a Database

In the database dialog, click Delete to remove an added database.

Query Database Data

After adding a database, the node uses the user's input question, combined with the added table structure and the built-in prompt in the User input box, to automatically generate and execute a SQL statement. The result is output to downstream nodes.

---

Parameters

Parameter	Description
Chat Memory	Enabled by default. When enabled, the User input box is bound to the Start node's query variable by default; when disabled, the User input box is hidden
Memory Window	Retains the last 50 turns of conversation context by default
Output Variable	Supports Array[String] type
Failure Retry	Supports automatic retry on execution failure
Exception Handling	Supports configuring exception branches to handle query failures and other errors

---

Typical Use Cases

Natural Language Database Query

Requirement: The user inputs "Query the product with the highest last month" and the system automatically generates SQL and returns the result.

Setup:

1. Create a database and import the business data table (e.g., sales records)
2. Add the database to the Conversational SQL node; enable chat memory
3. Bind the User input box to the Start node's query variable
4. The node automatically converts natural language to SQL and executes it; output is passed to a Direct Reply node

Start → Conversational SQL (natural language → SQL query) → Direct Reply

Multi-turn Follow-up Data Exploration

Requirement: The user first queries overall data, then progressively asks for details. The node needs to understand contextual references.

Setup:

1. Enable chat memory; increase the memory window size based on business complexity
2. The node passes conversation history to the AI, enabling it to understand references like "this one", "that one", "break it down further"

Start → Conversational SQL (with conversation history) → Direct Reply

FAQ

Q: What happens when chat memory is disabled?

The node no longer carries conversation history. Each query is an independent request — suitable for single-query scenarios that don't require context.

Q: How do I use the output in downstream nodes?

The output variable type is Array[String]. Reference it in a Direct Reply node or Model node using variable reference syntax.