AI Agent for Structured Data Extraction with LangChain & OpenAI

Unlock Robust Structured Data from LLMs with this AI Agent

This n8n workflow showcases an AI Agent built with LangChain and OpenAI, designed to transform natural language requests into clean, validated JSON output. It takes a text prompt, queries an OpenAI model, and then employs LangChain's powerful output parsing capabilities—including an auto-fixing mechanism—to ensure the data conforms to your predefined schema. This is crucial for anyone needing reliable, structured information from Large Language Models for their applications or automations.

Key Features & Benefits

• LangChain Orchestration: Leverages LangChain for advanced LLM interaction patterns, including robust output parsing.
• OpenAI Power: Utilizes OpenAI's chat models for sophisticated natural language understanding and generation.
• Schema Enforcement: Employs a StructuredOutputParser to define a target JSON schema, ensuring the LLM's output matches your exact requirements.
• Intelligent Auto-Fixing: Features an AutoFixingOutputParser that uses a secondary LLM call to attempt corrections if the initial output doesn't match the schema, significantly improving data reliability.
• Reliable JSON Output: Converts potentially inconsistent LLM responses into predictable, structured JSON, ready for use in APIs, databases, or further automation steps.
• Customizable & Adaptable: Easily modify the input prompt and JSON schema to suit a wide variety of data extraction tasks.

• An n8n instance (Cloud or self-hosted).
• OpenAI API Key with access to a suitable model (e.g., gpt-3.5-turbo, gpt-4).
• LangChain nodes enabled in your n8n instance (usually available by default in recent n8n versions).

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Configure both 'OpenAI Chat Model' nodes (one for the main query, one for auto-fixing) with your OpenAI API Key by selecting or creating the appropriate credentials.
4. In the 'Prompt' node, update the value field with your specific natural language query for data extraction.
5. Critically, examine the 'Structured Output Parser' node. Modify the jsonSchema parameter to accurately define the structure of the JSON data you intend to extract. If you expect a list of items, ensure your schema root is "type": "array" with the item schema nested under "items".
6. The 'LLM Chain' and 'Auto-fixing Output Parser' are interconnected. The auto-fixing parser attempts to repair outputs that don't conform to your schema using an additional LLM call, enhancing output quality.
7. Activate the workflow. Test with various prompts to ensure the structured output is accurate and meets your needs.