How to set up and customize your app?
Below is a step-by-step guide on how to set up and customize your app for optimal user interaction.
1. App overview
The app overview section displays the public link to your ZBrain app, provides the embed script for website integration, and shows a live preview of the chat interface. It allows users to manage basic settings like user info collection and provides quick navigation to configuration and performance sections.
Sharing your app
Once created, you'll find a link to your app on the overview page. Share this link with others to grant access to your app outside of the platform.
Adding a conversational widget to your website
Users can integrate a conversational widget into their website by pasting the provided code into the head section. This widget will appear at the bottom right corner of the page, allowing users to engage with the bot.
Collecting user information
Toggle the ‘Collect User Info’ option to prompt users to provide information when interacting with the bot. Users can specify the type of information to collect and provide a title for the user information form. After toggling, click ‘Save’ to apply the changes.
Editing appearance
Click on the 'Edit your appearance' button to access the appearance tab, where you can configure your bot's welcome message, app name, description, sample questions, logo, theme, and bot name. The preview panel on the right displays how your chatbot will appear to users in real-time. After making your desired changes, click 'Save and update preview' to apply all modifications.
2. Configuring your bot
Navigate to the ‘Configure your bot’ tab to access customization options that enhance the functionality and user experience of your application.
Orchestration method
Select your orchestration method: You can choose between using a knowledge base for information-driven responses or employing predefined Flows to guide interactions.
Adding resources For knowledge base: Include relevant knowledge bases your bot will refer to for information. Additionally, define the schema to structure the knowledge base for optimal retrieval and presentation of information. For Flow: If opting for Flows, integrate the necessary predefined Flows that your bot will follow during interactions to ensure smooth and guided user experiences.
Note: Knowledge Base orchestration requires configuring system instructions, adjusting bot settings, and adding golden questions for evaluation, while Flow orchestration only requires adding golden questions since conversation paths are predefined.
Adding instructions
Provide specific instructions or guidelines that the bot should follow during user interactions.
Click ‘Generate’ to create instructions by entering the details you want to update, then click ‘Create’ to finalize them.
Alternatively, click Library to manually create a new prompt or generate one automatically. Click ‘Save’ to store your prompt for future use, then click ‘Use’ to apply the saved prompt in the app.
To add a prompt from the Prompt Library, search for a template and select the desired prompt.
Evaluate bot settings
Define golden questions and responses to effectively assess and enhance your bot's performance.
Enter the question and corresponding answer in the provided fields.
Click ‘Add’ to include more golden questions for evaluation.
Bot settings
View and manage the current configurations of your bot, including model selection, response customization, and optimization settings.
Current settings
Check key parameters under bot settings, including:
Model name – The AI model used for generating responses.
Temperature – Controls response randomness; higher values create more varied outputs.
Content max token – Limits the number of tokens (words and characters) in generated responses.
Click ‘View All Settings’ to access additional parameters:
Top P (Nucleus sampling) – Controls diversity by considering only the most probable responses.
Presence penalty – Encourages topic variation by discouraging repeated words.
Frequency penalty – Reduces repetition by penalizing frequently used phrases.
Context max token – Defines how much past conversation is retained for context.
Response max token – Sets the maximum length of AI-generated responses.
Guardrails – Implements safety constraints to ensure appropriate responses.
Static input – Adds predefined text for consistent context.
Rerank – Improves search relevance by prioritizing the most relevant responses.
Follow-up conversation – Allows the model to recall past interactions within the same session.
Source – Enables the model to summarize documents in responses.
Editing model parameters
Click the pencil icon to edit settings.
Select the preferred AI model.
Adjust parameters using the provided controls:
Click ‘Load Presets’ and choose:
Creative – Generates imaginative and diverse responses.
Balanced – Maintains a mix of creativity and accuracy.
Precise – Focuses on delivering concise and factual responses.
Use sliders to configure:
Temperature – Adjusts response variability.
Top P – Filters response choices based on probability.
Presence penalty – Encourages diverse topics by discouraging repetition.
Frequency penalty – Reduces repetitive phrases.
Context max token – Determines how much past conversation the AI remembers.
Response max token – Sets the length limit for AI responses.
Enable or disable features as needed:
Reranking model – Enhances search result relevance.
Follow-up conversation – Enables the model to recall previous interactions.
Guardrails – Applies safety constraints to AI responses.
Static input – Adds static text for context consistency.
Source – Allows document summarization in responses.
Saving and reverting settings
Update settings – Click to save your customized configurations.
Default settings – Click to revert to the original system configurations.
Manual and auto configuration
Manual configuration: To make advanced changes, click the ‘Edit’ button in the Manual Configure box. (This option is for advanced users only.)
Auto configuration: Click ‘Start’ to let the system automatically optimize settings for your app. This process takes 10-15 minutes, and you can track configuration accuracy under ‘Bot Performance.’
After completing all the necessary adjustments and configurations, click the ‘Save’ button to apply and save your settings.
3. Bot performance
After configuring the bot, click ‘Test Bot Performance’ to check the bot's performance report.
After initiating the test, you can review the performance report generated from the test. Detailed report includes:
Accuracy of each question: See the accuracy score for each individual question, including the ZBrain answer and the correct answer provided by you.
Default settings in use: View the current settings used during the test.
Current version: Identify the version of the bot configuration used for the test.
If needed, you can adjust the accuracy by clicking the ‘+’ button in the accuracy column. This will make the ‘Edit’ button appear, allowing you to make changes.
Click the ‘Update’ button to adjust and save the accuracy.
To delete a specific version, click the ‘Delete this Version’ button in the version settings to remove it from the list.
Click the download icon to save the bot performance report.
4. Query history
The ‘Query History’ tab provides valuable insights into the interactions users have had with your bot. Here's how you can leverage this feature:
Navigate to the ‘Query History’ tab on your dashboard.
On the ‘Query History’ page, you will find the following information:
Session IDs: Each interaction session with the bot is assigned a unique session ID, allowing for easy tracking and reference.
User information: Users' names and email addresses associated with each session are displayed, providing context for the queries.
Prompt count: The number of prompts or interactions within each session is recorded, giving an indication of user engagement and interaction depth.
Session time: The duration of each session is logged, helping to analyze the duration of user engagements and overall bot usage patterns.
Viewing session details
Click on any session to explore the full conversation history. In the session view, you'll find:
The session name displayed prominently at the top
User information (if enabled in the “Collect User Info” settings)
A complete session timeline with timestamps
Sorting options: Newest first or Oldest first
Reviewing bot responses
Each bot response in the conversation includes several action options:
Copy output: Quickly copy the response for reference
View conversation log: Access a detailed breakdown of the exchange
Edit annotation reply: Modify the bot’s response for accuracy
Provide feedback: Use the thumbs-down icon to flag issues
Conversation log features
Clicking ‘Conversation log’ opens two key tabs:
1. Overview tab
Get a high-level summary of the interaction:
Conversation status
Total elapsed time
Tokens used (input and output)
Input query and bot response
Metadata including:
Model used
Temperature settings
Input/output token count
Number of chunks retrieved
Applied instructions
Click the ‘Monitor’ button to actively monitor your app
2. Context tab
Dive deeper into the information behind the response:
View all referenced chunks
Matching scores for each chunk
Click individual chunks to view full details
Editing annotations
To improve or correct a bot response:
Scroll to the feedback section beneath the query and response
Click the pencil icon
Enter your corrected version of the response
Click ‘Save’ to update the annotation
This comprehensive view of user interactions allows you to understand how users engage with your bot, identify common questions, analyze response quality, and continuously improve your bot's performance through annotation and feedback.
Last updated