Knowledge-Base

Knowledge base acts as a database to MEVA, where it can fetch information related to user queries in detail. This node helps in attaching the required document to the flow. Whenever a user asks for any information, MEVA understands the intention and triggers the kb node to fetch data for the user.

The kb node contains,

1. Search Type

   a. Tags: To enhance retrieval accuracy, MEVA should fetch documents containing matching tags, which serve as metadata to identify relevant content precisely.

   b. Documents: If you want to add only one document to the node, then you can select this radio button. A drop-down appears below, from where you can select the document name and MEVA will fetch data from that specific document.

2. Similarity Score: A similarity score in a knowledge base (KB) quantifies how closely a retrieved document or snippet matches a user's query. It is calculated using algorithms that compare the query with the available content, often based on factors like keyword overlap, semantic relevance, and contextual similarity. High similarity scores indicate that the content is highly relevant to the user's query, ensuring accurate and useful responses. The similarity score ranges from 0-1.

3. Pre Prompt

4. Post Prompt

For Example: FAQ documents related to IT Helpdesk. The Pre-Prompt can be,

            "You are an AI assistant designed to provide technical support. When a
            user asks about setting up their email on a smartphone, respond with a
            clear and friendly guide. Make sure to cover the basic steps and offer
            to provide further assistance if needed. Tailor the instructions for
            both Android and iPhone users."