Trust Center RAG Chatbot

How it Works / How to Build It

• SearchNode: This node takes the user's question as input and searches the "vellum-trust-center-policies" document index for relevant information. It limits the results to 8 documents and uses a combination of semantic similarity and keyword matching to rank the results.
• FormattedSearchResults: This node formats the search results into a readable template. It organizes the results by displaying the policy label and the corresponding text, separating each result with a line break for clarity.
• PromptNode: This node generates a response based on the formatted search results and the user's question. It uses a predefined prompt that instructs the AI to answer the question strictly based on the provided context from the security policies, including citations.
• FinalOutput2: This node outputs the final response generated by the PromptNode, which is the answer to the user's question.

What You Can Use This For

• Answering security-related inquiries from employees or clients.
• Providing quick access to specific security policies during audits.
• Assisting compliance teams in responding to security questionnaires.

Prerequisites

• Vellum account.
• Access to the "vellum-trust-center-policies" document index.
• Basic understanding of how to use Vellum's workflow builder.

How to Set It Up

• Clone the workflow template from the Vellum library.
• Ensure you have the necessary documents indexed in "vellum-trust-center-policies."
• Configure the SearchNode to point to your document index and adjust the query settings if needed.
• Set up the FormattedSearchResults node to format the output as desired.
• Customize the PromptNode to refine the AI's response style or content if necessary.
• Connect the nodes in the specified order: SearchNode >> FormattedSearchResults >> PromptNode >> FinalOutput2.
• Test the workflow by inputting sample questions to ensure it returns accurate and relevant answers.