DB Fiddle

By Status200 Ltd

Overview

The DB Fiddle plugin lets users run SQL against a disposable sandbox environment, directly from ChatGPT and Codex. Create database schemas, load sample data, execute queries, and inspect real database results without connecting to a production database or external system. The sandbox provides the perfect place for ChatGPT and Codex to iterate on queries and safely debug SQL, reproduce database issues, validate queries and experiment with any of the supported database engines. Supported database engines: PostgreSQL, MySQL, MariaDB, SQLite and more. Note: A db-fiddle.com PRO subscription is required to use this plugin.

Tools

list-engines

ChatGPT

List the available database engines and their version selection IDs. Use this tool when the assistant needs to choose a supported engine/version before executing SQL, or when the user explicitly asks which database engines or versions are available. Do not use this tool as the only step for SQL-building, testing, debugging, validation, optimisation, or demonstration requests. After identifying the appropriate engine/version, use the relevant SQL execution tool unless the user only asked for engine availability. Return or summarise only the engine names and version IDs needed for the user’s request. Avoid listing every engine/version when a narrower answer is sufficient.

run-sql

ChatGPT

Run SQL against a selected database engine/version and show the execution UI/results to the user. Use this tool when SQL execution itself is the user-facing deliverable. Use this tool when the user explicitly asks to: - run SQL - execute SQL - test a query and show the result - demonstrate SQL with visible output - debug a query using the returned database error/result - create a fiddle/example the user can inspect - compare visible result sets Do not use this tool merely to check the assistant’s assumptions, confirm database-specific behaviour, verify syntax before answering, or inspect intermediate results privately. For those cases, use run-sql-silent. If the user has not provided schema or sample data and visible execution is requested, create a minimal representative schema and sample rows. After execution, do not repeat the full SQL or full result set. The UI already displays it. Summarise whether it succeeded and what the result demonstrates.

run-sql-silent

ChatGPT

Run SQL privately against a selected database engine/version without showing the execution UI/results to the user. Use this tool whenever SQL execution is needed to verify, support, or improve the assistant’s answer, but the user did not explicitly ask to see a visible result table, fiddle, execution UI, or database error output. This tool MUST be used for questions about database-specific behaviour that can be tested with a minimal query, including: - function semantics, such as numeric, string, date/time, JSON, aggregate, or window functions - operators and comparisons - NULL behaviour - type coercion and casting - collation and sorting - date/time arithmetic - dialect-specific syntax or return values - edge cases involving literals, boundary values, precision, rounding, overflow, or invalid inputs - “what does this return?” questions - “does database X do A or B?” questions Use this tool even if the question sounds conceptual, documentation-style, simple, or answerable from memory. Database edge cases are easy to misremember; verify them when executable. Examples that require this tool: - “Does PostgreSQL concatenate NULL as an empty string?” - “What does MySQL return for '10' < 2?” - “In SQLite, what does date('2024-02-31') return?” - “Does SQL Server COUNT(column) include NULLs?” - “What does Oracle return for an empty string?” Use run-sql instead only when the user explicitly asks to see the execution output, result set, inspectable fiddle, or error message. Do not use this tool when: - SQL execution is impossible without missing schema/sample data and no useful minimal example can be created - the question is purely about writing, formatting, or explaining standard SQL without needing verification - the user explicitly asks for visible execution output After using this tool, answer with the verified conclusion and explanation. Do not mention private execution unless it is useful context.

Capabilities

Interactive

Example Prompts

Click any prompt to copy it.

Links

Documentation Privacy Policy

Similar Apps