AI Agent for Reliable Structured Data Extraction with LangChain & OpenAI

Unlock Reliable Structured Data from LLMs with this AI Agent

This n8n workflow acts as a sophisticated AI Agent designed to extract structured data (JSON) from Large Language Model (LLM) responses to your natural language prompts. Its key strength lies in its ability to ensure the output conforms to a predefined schema, even if the LLM's initial response isn't perfectly formatted. The agent intelligently uses a primary LLM for generating the initial data and then employs an auto-fixing mechanism, powered by a secondary LLM and a defined JSON schema, to correct any malformed outputs. This significantly increases the reliability of getting usable, structured JSON, making it invaluable for automating data entry, content enrichment, and feeding structured information into other business systems or databases.

Key Features & Benefits

• AI-Driven Data Structuring: Leverages the power of OpenAI's language models via LangChain to understand complex natural language prompts and generate data in a structured format.
• Automated Output Correction: Features LangChain's Auto-fixing Output Parser, which utilizes a secondary LLM to intelligently correct and reformat the initial LLM output to match your specified JSON schema.
• Guaranteed JSON Schema Adherence: Increases the reliability of obtaining well-formed JSON outputs, crucial for robust downstream automation and data integration.
• Flexible Natural Language Input: Processes user-defined prompts, allowing you to ask for diverse and complex information.
• Customizable Output Schemas: Easily define your desired output structure using JSON Schema within the 'Structured Output Parser' node, tailoring the agent to your specific data needs.
• Seamless n8n Integration: Built natively within n8n, enabling straightforward connection to hundreds of other applications and services for end-to-end process automation.

• An n8n instance (Cloud or self-hosted).
• OpenAI API Key with access to suitable models (e.g., gpt-3.5-turbo, gpt-4). You'll need to configure credentials for two 'OpenAI Chat Model' nodes in the workflow.
• A basic understanding of JSON Schema to define your desired output structure in the 'Structured Output Parser' node.

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Locate the two 'OpenAI Chat Model' nodes:
   • One connected to the 'LLM Chain' node (this is for the primary data generation).
   • Another connected to the 'Auto-fixing Output Parser' node (this is for correcting the output). Configure both with your OpenAI API Key and select your preferred models.
4. In the 'Prompt' node (a Set node), modify the input value to specify the natural language query you want the AI Agent to process.
5. In the 'Structured Output Parser' node, carefully define the jsonSchema parameter. This schema dictates the exact JSON structure the agent will try to produce and validate against. The provided example requests US states, their largest cities, and populations.
6. The 'LLM Chain' node is pre-configured to use the primary LLM and the 'Auto-fixing Output Parser'. The 'Auto-fixing Output Parser' itself uses the 'Structured Output Parser' for schema definition and the second LLM for attempting fixes.
7. Activate the workflow. You can test it by clicking 'Execute Workflow' on the manual trigger.