Welcome to WorkerBot!
17 January 2025
WorkerBot is an experimental application that showcases the potential of AI in enhancing worker support systems. It is a Proof-of-Concept collaboration between University of Manchester and FiftyEight (https://fiftyeight.io/), and funded by the UK Research and Innovation (UKRI) Impact Acceleration Account (IAA).
Getting Started: To access WorkerBot, please email enduringnet at gmail dot com.
Navigating the Application: Once you are in the application, you can navigate through it using the sidebar. Here's what you can find:
- Settings: Customise your app settings here.
- Q&A (Chat Interface): Start here for your main interactions. This is the default landing page where you can ask questions and receive guidance.
- About: Learn more about WorkerBot and its development.
A. Q&A Page Overview
The Q&A page is your main interface for interacting with WorkerBot. Here’s how to navigate and use its features:
Control Buttons
At the top of the Q&A page, you will find three control buttons:
- Save Chat (💾): Click to save your entire conversation history and trigger the feedback system.
- New Chat (🔄): Click to clear all current conversation history and start a new chat session.
- Voice Input (🎤): Click to enable the voice input feature, which supports multiple languages and displays a transcription preview.
Ask Questions
You can ask questions in two ways:
1. Text Input: Type your query into the text box at the bottom of the page. Press Enter or click the send button to submit your question to WorkerBot.
2. Voice Input:
- Click the Voice Input (🎤) button at the top of the page.
- In the voice input interface, click the microphone icon to start recording.
- Click the microphone again to stop recording and send your query to WorkerBot.
Conversation Display
Messages are displayed in chronological order and each message is visually separated for clarity:
- Your messages are prefixed with "You:"
- WorkerBot’s responses are prefixed with "Assistant:"
After Your Chat
After you finish your chat, click the Save Chat (💾) button to save the conversation history and trigger the feedback system at the bottom of the chat. Please note that saved data is not accessible to users.
B. Settings Page Overview
The Settings page is organised into three main tabs: Model Settings, Language & Persona, and Knowledge Database.
1. Model Settings Tab
- Model Selection: Choose a language model (LLM) from the available options to process your queries. Available models include:
- Mistral-Small (22B)
- Llama3.3 (70B)
- Llama3.2 (3B)
- Llama3.1 (8B)
The number (e.g., 22B) indicates the size of the model. Larger models typically provide better outputs but require more time to process queries.
- Temperature Setting: Adjust the Temperature to impact the variability of responses:
- A lower Temperature results in more consistent and predictable answers.
- A higher Temperature allows for more varied and creative responses.
2. Language & Persona Tab
- Language Selection: Choose the language in which WorkerBot will respond. This setting only affects the output; inputs can be in a different language. The default setting is English with a friendly personality.
3. Knowledge Database Tab
- Context Data Selection: Select specific context data from the available knowledge bases to enhance the relevance of responses by defining the scope of available information.
- Note: The embedding model currently supports processing (knowledge base) in English only.
- Vector DB Mode: This feature is not for user interaction. It allows for the creation of new databases using text files present in the knowledgebase folder in the root directory and lets administrators choose from various existing knowledge databases.