ChatFAQ Admin¶
The ChatFAQ admin will be a node application that allows you to manage your chatbot’s relevant data from the browser.
Dashboard¶
The ChatFAQ Admin Panel provides a comprehensive interface for accessing various aspects of the platform. The dashboard offers an overview of key components and their status. Key features include:
FSMs (Finite State Machines):
Displays a list of existing FSMs.
Shows the FSM name and creation date.
Retrievers:
Lists all enabled retrievers.
For each retriever, shows:
Enabled/Disabled status (with a toggle switch)
Model name
Retriever type
Includes a “Reindex” button for each retriever.
If no retrievers are available, a message indicates this.
LLMs (Language Models):
Lists all enabled language models.
For each LLM, shows:
Enabled/Disabled status (with a toggle switch)
LLM type
LLM name
If no LLMs are available, a message indicates this.
Widgets:
Displays a list of all widgets.
For each widget, shows:
Domain
FSM (Finite State Machine) definition
If no widgets are available, a message indicates this.
Labeling¶
The Labeling section of the ChatFAQ admin tool facilitate the management and improvement of the chatbot’s interactions.
Initial screen: Conversation Search and Filtering¶
Search bar : Locate specific conversations of interest based on their content or title.
Date Range Selection: Filter conversations by their date range.
Conversation Status Filter: To efficiently manage your review process, there are three statuses you can filter by:
Reviewed: Conversations that all Question & Answer (Q&A) units have been fully reviewed.
To Review: Conversations where no Q&A units have been reviewed yet, indicating they are pending your analysis.
Under Review: Conversations where some Q&A units are currently under review, indicating an ongoing evaluation process.
Detailed Labeling Screen¶
Once you select a conversation on the initial Labeling screen, you will be directed to the detailed labeling screen. This screen is designed to evaluation and tagging of each interaction.
Left side: Conversation Layout and Content¶
General Information: On the left side of the screen, you will see the complete conversation. Each conversation includes its title, the RAG (Retriever-Augmented Generation) it belongs to, and the start date of the conversation.
Conversational Bubbles:
Right Side (Question): Represents the user’s questions or comments.
Left Side (Answer): Displays the answers generated by the AI chatbot. These responses are the primary focus of your evaluation.
Q&A Blocks: Each pair of question and answer forms a Q&A block, which is the key element for evaluation. Select one pair to start the evaluation and tagging process.
Right side: Evaluation and Tagging Process¶
Evaluation of Knowledge Items: Determine if the knowledge items used by the retriever are appropriate or if there is a need to add more relevant information.
Feedback and Rating:
Rating from 1 to 5: Provide a numerical rating for each chatbot response, based on criteria defined by the business.
Text Feedback: Specific feedback, including suggestions for alternative text or necessary revisions in the chatbots response.
User Feedback: Check whether the user has given positive or negative feedback and review any additional comments they might have provided.
Data¶
Thanks to this module, you will access databases, knowledge items, and search or add more data in this tab to manage and enrich your chatbot’s informational resources.
Knowledge Base¶
Knoledge Item¶
You could access the knowledge items in your knowledge base to examine the splitter units of information that the RAG (Retrieval-Augmented Generation) system retrieves as part of its response generation process.
Intents¶
Optimize chatbot understanding by managing existing intents and exploring user-based intent suggestions.
Existing Intents¶
Suggested Intents¶
AI Configuration¶
Widget Configuration¶
SDK Visualizer¶
Stats¶
It is currently under development.