How to find the chunks on which the chatbot bases its answers in the document analysis bubble (RAG)?

Documentation: Accessing Document Snippets for Chatbot Responses

Our software incorporates an advanced feature that allows users to access document snippets on which the chatbot bases its responses. This feature is essential for understanding how the chatbot generates its answers and what information it relies on. Here is a detailed guide on how to use this feature.

Step 1: Access the Testing Section

  1. Navigation: Start by heading to the tab on the left side of the user interface, titled “Testing”. This section is specifically designed to test interactions with the chatbot and examine how it responds to different questions.

Step 2: Ask a Question

  1. Interacting with the Chatbot: Once in the Testing section, ask a question to the chatbot using the provided input field. This step is crucial as it triggers the chatbot’s analysis process to generate a response.

Step 3: Open the Data Tab

Data Access: After asking your question, navigate to the “Data” tab located on the right side of the interface. This tab is usually open by default, but if not, you can open it by clicking on its icon “show conversation data” (see below).

Capture d’écran 2024-03-25 à 22.53.08

Step 4: Exploring Sources in the AI Section

  1. **AI Section > Sources ** find the section titled “AI”. This section contains detailed information on the artificial intelligence processes involved in generating responses.
  2. Accessing Sources: In the AI section, locate the subsection “Sources”. Here, you’ll find a list of various information sources provided to the large language model (LLM), such as ChatGPT or others, which can be then used to enrich the chatbot’s knowledge base and enable it to respond informedly to your questions.


By following these steps, you will be able to access and understand the document snippets and information sources on which the chatbot bases its responses. This feature provides valuable transparency into the chatbot’s internal workings and allows users to verify the reliability and relevance of the information provided.