AI RAG Agent for Automated Citation Generation with OpenAI

Unlock Accurate, Verifiable Content with this AI Agent

This n8n workflow functions as an intelligent AI Agent, leveraging OpenAI Assistants enhanced with Retrieval Augmented Generation (RAG). It connects to your pre-configured OpenAI Assistant that has access to a vector store of your documents. When you ask a question, the agent not only provides an answer based on the information in those documents but also meticulously extracts and formats citations, ensuring the generated content is traceable and verifiable.

This agent is a powerhouse for anyone needing to generate reliable, sourced information. It automates the tedious process of finding and referencing source materials, allowing you to focus on creating high-quality content or making informed decisions based on accurately cited data.

Key Features & Benefits

• AI-Driven RAG: Utilizes OpenAI Assistants with built-in file retrieval to access and process information from your custom knowledge base (vector store).
• Automated Citation Generation: Intelligently identifies and extracts citation details (like filenames) corresponding to the information used in the response.
• Enhanced Verifiability: Provides clear links or references back to the source documents, building trust and transparency in AI-generated content.
• Customizable Output: Formats citations within the text, with options to customize the style (e.g., Markdown links) and even convert the final output to HTML.
• Streamlined Research & Content Creation: Drastically speeds up the process of generating reports, articles, or documentation that require sourcing.
• Improved Accuracy: Grounds AI responses in your specific documents, reducing the likelihood of irrelevant or incorrect information.
• Interactive Testing: Includes an n8n Chat Trigger for easy testing and interaction directly within your n8n instance.

• An n8n instance (Cloud or self-hosted).
• An OpenAI API Key with access to the Assistants API (v2 recommended) and File API.
• An existing OpenAI Assistant ID. This Assistant must be pre-configured in your OpenAI account with file retrieval (vector store) enabled and your relevant source files already uploaded to it.

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Configure the 'OpenAI Assistant with Vector Store' node: Select your OpenAI API credential and paste your pre-configured OpenAI Assistant ID into the 'Assistant ID' field.
4. The 'Get ALL Thread Content' and 'Retrieve file name from a file ID' HTTP Request nodes are pre-configured for OpenAI. Ensure your selected OpenAI credential has permissions for threads and files endpoints.
5. Examine the 'Finnaly format the output' Code node. This is where the magic happens: it replaces citation placeholders (e.g., 【13†source】) with formatted filenames (e.g., _(filename.pdf)_). You can customize the output string here, for instance, to create Markdown links if your filenames can be mapped to URLs.
6. The 'Optional Markdown to HTML' node is disabled by default. If you need HTML output, enable it and ensure its input ({{ $json.output }}) correctly references the output of the 'Finnaly format the output' node.
7. Activate the workflow. You can test its citation generation ability using the 'Create a simple Trigger to have the Chat button within N8N' node by sending messages via the n8n chat interface.