Make OpenAI Citation for File Retrieval RAG

For OpenAI, this automated workflow retrieves and formats citations from files, ensuring accurate sourcing in responses. It aggregates data from multiple threads, splits content for detailed processing, and allows for Markdown or HTML output. This enhances the reliability of information provided by the OpenAI assistant, streamlining the citation process and improving content presentation.

7/8/2025
19 nodes
Complex
urxrtgxxlobzwpvxnmxs3c9l1wqdwwf5manualcomplexaggregatelangchainsticky notesplitoutmarkdownadvancedapiintegration
Categories:
Complex WorkflowManual Triggered
Integrations:
AggregateLangChainSticky NoteSplitOutMarkdown

Target Audience

This workflow is designed for:
- Developers seeking to automate citation retrieval from OpenAI's vector store.
- Researchers who need to ensure proper citation of sources in their work.
- Content Creators looking for efficient ways to integrate citations into their materials.
- Data Analysts who require accurate references from AI-generated content.
- Educators wanting to provide students with tools for proper source attribution.

Problem Solved

This workflow addresses the challenge of retrieving accurate citations and sources from OpenAI's assistant when generating content. Often, the assistant may not provide all necessary citations, leading to incomplete or inaccurate references. By automating the process, users can ensure that all relevant citations are retrieved and formatted correctly, thus enhancing the reliability of their work.

Workflow Steps

  • Trigger the Workflow: The process starts with a manual trigger, allowing users to initiate the workflow whenever needed.
    2. OpenAI Assistant Interaction: The workflow utilizes an OpenAI assistant integrated with a vector store to perform content retrieval.
    3. Retrieve Thread Content: An HTTP request fetches all messages from a specific thread, ensuring that all relevant information is collected.
    4. Split Messages: The workflow splits the retrieved messages into individual iterations, allowing for detailed processing of each message.
    5. Extract Content and Citations: Each message's content and associated citations are separated for further handling.
    6. File Name Retrieval: For each citation, the workflow makes an additional HTTP request to obtain the file name associated with the citation ID.
    7. Regularize Output: The output is structured to contain essential information: citation ID, file name, and text.
    8. Aggregate Data: All collected data is aggregated to facilitate a single request for formatting.
    9. Format Output: The final step involves formatting the output using Markdown or HTML, ensuring citations are properly linked and presented.

    • Customization Guide

    To customize this workflow:
    - Modify the OpenAI Key: Ensure that the OpenAI API key is correctly configured in the credentials section.
    - Adjust Formatting: Users can edit the JavaScript code in the 'Finnaly format the output' node to change how citations are formatted in the output (e.g., switching between Markdown and HTML).
    - Add More Nodes: Additional nodes can be included for more complex processing or to integrate with other APIs.
    - Change Trigger Type: Users can adapt the trigger type to suit their needs, such as switching to an automated trigger based on specific events.