Automated AI SQL Generator and PostgreSQL Error Debugger - n8n Workflow

• Database developers looking for an AI-assisted SQL sandbox.


• Technical teams requiring automated, validated SQL generation.


• n8n users seeking advanced examples of AI integration and error handling loops.


• Anyone who wants a ready-to-use n8n template for interacting with databases via natural language.

This advanced n8n workflow template solves the significant challenge of manually validating AI-generated SQL. By integrating large language models (LLMs) like GPT (via OpenAI or OpenRouter) directly with a PostgreSQL database, this n8n automation creates a highly efficient and self-debugging SQL development tool. The core value of this n8n workflow lies in its built-in error handling loop. If the initial SQL execution fails in the sandbox, the system automatically captures the error details via a dedicated n8n node (ExecuteAIresult), sends the error back to the AI for self-correction, and re-attempts execution. This powerful loop ensures the output is always executable, making this a highly robust and reliable n8n workflow.

The entire process is initiated by a specialized n8n trigger and moves through several conditional steps to manage AI interaction and database execution.

1. Trigger and Setup: The n8n workflow begins with the When chat message received n8n trigger, capturing the user's natural language request and crucial configuration parameters, such as the desired aiProvider and the autoErrorFixing setting. The localVariables n8n node then sets up the strict, detailed instruction prompt necessary for directing the AI to act as a 'Senior PostgreSQL developer,' ensuring structured SQL output.


2. AI Orchestration: The workflow branches based on the aiProvider. If OpenAI is selected, the n8n template uses the Assistant API for seamless history management. If OpenRouter is chosen, the Simple Memory n8n node is used to maintain conversational context.


3. SQL Generation: The primary AI n8n nodes (OpenAIMainBrain or OpenRouterAgent) generate the complete SQL script, which includes schema creation (CREATE TABLE), mock data insertion, and the final solution query.


4. Execution and Error Handling: The ExecuteAIresult PostgreSQL n8n node attempts to execute the generated code. If successful, the result is captured. If an error occurs, the flow checks the AutoErrorFixing parameter.


5. Debugging Loop: If auto-fixing is enabled, the GenerateErrorPrompt n8n node constructs a specific prompt detailing the execution error. This prompt is sent back to the AI for refinement, creating a loop. The IsMaxAutoErrorReached n8n node acts as a safeguard, limiting the loop iteration count (to 4) to prevent indefinite execution, ensuring this powerful n8n template remains cost-effective.

To deploy this powerful n8n workflow, follow these steps:

1. Import: Copy the provided JSON and import it into your self-hosted or cloud n8n instance.


2. PostgreSQL Credentials: Locate the ExecuteAIresult n8n node and set up your PostgreSQL connection credentials. This database acts as the SQL sandbox where the generated code is executed.


3. AI Credentials: Update the credentials for both OpenAI and OpenRouter n8n nodes, depending on which provider you plan to use. Ensure your credentials have the necessary permissions.


4. Configuration: The initial When chat message received n8n trigger requires specific input data (e.g., sessionId, chatInput, aiProvider). Ensure your application or frontend supplies these parameters correctly when invoking this n8n workflow via its webhook ID.


5. Testing: Execute the n8n workflow with a test SQL query request to verify that the AI can generate, execute, and automatically fix errors.

When chat message received (LangChain Chat Trigger): The starting n8n trigger that captures user input (natural language prompt) and required configuration parameters like aiProvider, model, and autoErrorFixing.
localVariables (Set): Initializes the core instructional prompt for the AI, defining its role as a senior PostgreSQL developer and establishing the strict SQL formatting rules required by this n8n template.
OpenAIMainBrain (OpenAI): The key n8n node for OpenAI users, interacting with the OpenAI Assistant API. It manages message threads and leverages existing assistants or creates a new one based on the model specified in the n8n trigger data.
OpenRouterAgent (LangChain Agent): The main n8n node for users opting for OpenRouter. It uses the Simple Memory n8n node to maintain the session history for the AI, adapting the n8n workflow for different LLM providers.
ExecuteAIresult (Postgres): The critical database n8n node. It attempts to execute the full SQL script generated by the AI. It is configured to continue to the error output if execution fails, facilitating the self-correction loop.
GenerateErrorPrompt (Code): When execution fails, this custom n8n node generates a new, highly specific prompt. This prompt includes the original user request and the technical PostgreSQL error message, instructing the AI to fix the query.

• IsMaxAutoErrorReached (If): A flow control n8n node that counts the number of times the error fix loop has run (limited to 4 attempts) to prevent costly, infinite execution cycles in this n8n workflow.