Skip to content
Nappai Documentation

libSQLRetrieverTool

This component lets you pull information from a libSQL or Turso database and turn it into a tool that can be used by other parts of your Nappai workflow. It’s especially handy when you want to ask a language model to fetch data from a database without writing code yourself.

⚠️ DEPRECATION WARNING

This component is deprecated and will be removed in a future version of Nappai. Please migrate to the recommended alternative components.

How it Works

The libSQLRetrieverTool connects to a libSQL/Turso database using the connection details you provide in Database Info. It then runs the SQL query you write in SQL Query Template. The query must return a column named text; the content of that column is turned into a vector using the Embedding you supply. Finally, the tool is packaged so that other components (like a language model) can ask it to retrieve the most relevant text snippets based on a user’s question.

Inputs

  • Database Info: Connection details for the libSQL/Turso database you want to query.
  • Embedding: The embedding model or method that will convert the retrieved text into a vector for similarity search.
  • Metadata Schema: The schema that describes the structure of the data you’re retrieving, used by the SelfQueryRetriever.
  • SQL Query Template: A SQL statement that returns a column named text. You can use {[input]} placeholders to insert dynamic values.
  • Tool Description: A short description of what the tool does, used by the language model to understand its purpose.
  • Tool Name: The name you want the tool to have when it appears in the model’s tool list.

Outputs

  • Tool: A BaseTool object that can be passed to other components (e.g., a language model) to perform database queries on demand.

Usage Example

  1. Set up the component

    • Drag the libSQLRetrieverTool onto your canvas.
    • Fill in Database Info with your libSQL/Turso credentials.
    • Choose an Embedding model (e.g., OpenAI’s text-embedding-ada-002).
    • Define the Metadata Schema that matches your database tables.
    • Write a SQL Query Template such as: 
      SELECT text FROM articles WHERE category = '{[category]}' LIMIT 5;
    • Provide a clear Tool Description like “Retrieves recent articles about a given category.”
    • Name the tool “ArticleFetcher”.

  2. Use the tool in a flow

    • Connect the output of a user‑input component to the {[category]} placeholder in the query template.
    • Pass the resulting Tool to a language model component that can call it when the user asks for articles on a topic.

  3. Run the workflow

    • The model will call the tool, the tool will query the database, embed the results, and return the most relevant text snippets.
  • SelfQueryRetriever – Builds on top of this tool to let a language model ask questions that automatically trigger database queries.
  • Embedding – The component that provides the embedding model used here.
  • Database Connector – A more general connector for other database types.

Tips and Best Practices

  • Always return a column named text; otherwise the tool will not work.
  • Test your SQL query directly in the database first to ensure it returns the expected results.
  • Keep the query simple and limit the number of rows to avoid long response times.
  • Use placeholders carefully; they should match the names you use in the input fields.
  • If you need to filter by multiple parameters, add more placeholders and provide the corresponding inputs.

Security Considerations

  • Store database credentials securely; avoid hard‑coding them in the workflow.
  • Restrict the database user’s permissions to only the tables needed for the query.
  • Validate any user‑supplied input that gets inserted into the SQL template to prevent injection attacks.
  • Monitor query logs for unusual activity and set up alerts if needed.