This workflow automates the process of retrieving and formatting citations from files stored within an OpenAI vector store, enhancing the accuracy and traceability of AI-generated content. It begins with a trigger for a chat button, allowing users to initiate the process within n8n. Once triggered, the workflow calls an OpenAI assistant integrated with a vector store containing relevant files.

The workflow retrieves all messages from a specific thread to ensure comprehensive citation gathering, as OpenAI’s API does not include citations by default. It then splits the thread content into manageable chunks: individual messages, message content, and separate citations to facilitate detailed processing.

For each citation, the workflow fetches associated file metadata, extracts the text, and normalizes the data to prepare it for formatting. All the citations and extracted texts are then aggregated into a unified output, where the final step carefully replaces citation references within the text with formatted Markdown links or HTML tags, making references clear and navigable.

This setup is particularly useful for scenarios requiring accurate sourcing in AI-generated reports, academic writing, or document management, ensuring citations are correctly matched, formatted, and easily accessible.

Node Count	11 – 20 Nodes
Nodes Used	@n8n/n8n-nodes-langchain.chatTrigger, @n8n/n8n-nodes-langchain.memoryBufferWindow, @n8n/n8n-nodes-langchain.openAi, aggregate, code, httpRequest, markdown, set, splitOut, stickyNote