Changelog

July 9, 2024

Enhanced Integrations and Improved Features

πŸš€ New Features
  • Source Citation Toggle: ou can now select which sources you want to be cited or not. With this feature, you can ensure that customers won’t have direct access to some of the sources but can still ask questions related to them.
  • Facebook Integrations: We are grateful to introduce a few highly anticipated integrations β€” WhatsApp, Instagram, and Messenger. Now you can connect your bots to any of these platforms and provide your users the ability to chat with either AI or human agents.
  • Make Integration New Action: Added a new action that allows users to use any GPT-trainer API along with Make.com, enhancing your automation capabilities.
πŸ’‘ Improvements
  • Source Overview: Now, clicking on a source will open a sidebar with a detailed explanation of it, where you can access and edit the metadata.
  • Make/Zapier Query-Response Action: You can now receive session UUIDs in the response, which lets you revisit the same session on separate Zapier requests, thereby retaining conversation history.
🐞 Bug Fixes
  • Long Message AI Responses: Fixed an issue where some long messages in different languages did not trigger AI responses correctly.
June 28, 2024

Transformative Updates: New Tools and Enhancements

πŸš€ New Features
  • GDPR Compliant: We are pleased to announce that our app is now GDPR compliant. This means we adhere to strict data protection and privacy regulations, ensuring that user data is handled with the highest level of security and transparency. Our commitment to GDPR compliance enhances our app’s reliability and trustworthiness, making us a better choice for users who prioritize data privacy.
  • XML File Upload: Expanding our existing sitemap functionality, you can now upload XML files directly as sources. Simply select the Website Sitemap option in the Add Sources section and choose your sitemap file.
  • Make.com Integration: Connect your chatbots with Make.com, an automation tool that allows seamless integration with thousands of apps. Our app is currently under review for full authorization on this platform, but it is already functional. Try it out here.
  • Inbox Search: Enhance your filtering capabilities by searching sessions based on UUIDs and their metadata. Metadata includes form fields, variable states, authentication data, and more, in addition to date-based filtering.
  • Image Generator Agent: Developed in partnership with Civitai. Currently in beta, this exciting feature allows you to generate your own images directly inside GPT-trainer! We are continuously expanding the capabilities of this agent. If you have particular use cases you'd like to present to us for fine-tuning, tell us via Discord or email at here.
πŸ’‘ Improvements
  • Widget Auto Open on Mobile Devices: The chat auto open feature will now only work on large screens, excluding mobile devices, for better user experience.
🐞 Bug Fixes
  • Empty Responses Messages: Fixed a bug where AI messages weren't generated successfully, resulting in empty response bubbles.
June 20, 2024

Enhancing User Interaction: New Forms and Response Actions

πŸš€ New Features
  • Custom Forms: Our latest feature for collecting custom information is now available. Now, you can collect data from your users either conversationally or via a built-in form! You can create personalized forms with various options, including text fields, checkboxes, and multi-option selections. Additionally, we have introduced a new webhook that triggers on form submission, sending email notifications, and integrating seamlessly with Zapier. This feature can be accessed by navigating to Settings / Appearance > Form section.
  • Response Actions: Enhance user engagement by enabling response actions. This feature allows users to vote on AI responses, providing valuable feedback on which responses they liked or disliked. Additionally, users can easily copy responses. You can find this feature in the Settings / Appearance > Conversations section. Votes on AI responses are stored in conversation metadata and can be exported for analysis via Inbox > Export > Chat History.
πŸ’‘ Improvements
  • Chat Appearance in Settings/Appearance: The chat settings have been updated to display all enabled features for your bot, allowing you to preview your chat widget's appearance in real-time.
  • Chat Redesign: Implemented a slight redesign of chat buttons and secondary action.
  • Inbox UI Improvements: The font size and overall UI in the inbox have been optimized for better visual clarity and usability.
  • AI Model Upgrades: All models have received an upgrade to improve default memory for chat history.
June 13, 2024

Chatbot Time Zone Awareness and Custom Message Monitoring

πŸš€ New Features
  • Chatbot Time Zone: Our chatbots can now understand and provide the current time based on the specified time zone. In Settings / Appearance > Conversations select your preferred time zone. You can design custom chatbot interactions that are dependent on the time of day!
  • Custom Messages Monitoring: The Human Escalation Agent has been upgraded with an innovative feature that allows you to monitor for specific messages output by the AI. If the AI generates a response similar to one of your monitored messages (such as "Sorry, I am not sure."), you’ll receive an in-app notification. This enables you to quickly access the relevant chat, review the AI’s response, and step in to take over the conversation with the user if the need arises. This is particularly useful for identifying knowledge gaps in your chatbot!
June 7, 2024

Enhanced Token Management and New Spam Defense Agent

πŸš€ New Features
  • Spam Defense Agent: Enhance the security of your chatbots with the newly upgraded Spam Defense Agent. This unique agent tracks and identifies spam messages directed at your bots without consuming your message credits. We'll be continuing to improve its accuracy and efficiency over time.
  • Advanced Hallucination Feature: Located in user-facing Agents'Knowledge tab this feature serves as an additional defense against LLM hallucinations (made-up responses). It assigns a relevance score for all retrieved chunks if your AI Agent is Retrieval Augmented Generation (RAG) enabled. If it is not confident enough about the source information, the AI Agent will output a fixed pre-canned response instead of forcefully generate one risking hallucination.
  • Token Distribution Options: Introducing a flexible option for User-Facing agents o customize token distribution for each part of the LLM call. This feature allows you to reserve a specific proportion of tokens, optimizing performance based on your specific use case. The total token limit is determined by the LLM you select, but fine-tuning is now in your hands.
πŸ’‘ Improvements
  • Sources Shift Select: You can now select multiple sources by holding down the Shift key.
  • Multi Source Download: You can now download multiple sources from your Sources library simultaneously.
  • Bubble Integration: We added a new guide for integrating chatbots with the Bubble website builder.
  • Webflow Integration: We added a new guide for integrating chatbots with Webflow websites.
  • Chatbot Font Styles: We have revamped the font loading logic for chatbots to enhance performance. The new system ensures faster font loading times, providing a smoother user experience.
  • We have revamped the font loading logic for chatbots to enhance performance. The new system ensures faster font loading times, providing a smoother user experience.
🐞 Bug Fixes
  • Email Letters: Resolved an issue where users did not receive emails from GPT-trainer on certain email clients like Outlook. This fix ensures reliable email delivery across all platforms.
May 31, 2024

Enhanced Security and Compliance with New Features

πŸš€ New Features
  • Disclaimer Popup: Introducing a powerful new feature to enhance compliance within your chatbots. You can now add a disclaimer popup that requires users to accept the terms before they can begin chatting. This is particularly beneficial for businesses focused on regulatory compliance and user consent. Each session will log the acceptance date, ensuring you have a clear record of user agreement. Access this feature in the Settings / Appearance tab.
πŸ’‘ Improvements
  • Authorization Security: We have significantly increased the security for authorization flows, ensuring even more robust protection for our users accounts and their data.
May 28, 2024

Introducing Chatbot Templates and Enhanced Usability

πŸš€ New Features
  • Chatbot Templates: We’re introducing a new way to create chatbots. Now you can use one of our new pre-designed templates instead of starting from scratch! These templates will help you create a production-ready chatbot with minimal effort, and you can update them later as needed.
πŸ’‘ Improvements
  • Chatbot Menu: Redesigned the chatbot menu for better usability. Instead of a dropdown, you now have a full dashboard where you can manage your chatbots more efficiently.
  • Improved Message Credits Visibility: If an AI Agent is using any LLM, it will now display the exact message credit amount used per user interaction inside the Agent’s AI Model badge.
🐞 Bug Fixes
  • Function Calling Error Handling: Fixed an issue where chatbots were not receiving any response if Agent functions were stuck or had issues with getting a response.
  • Inbox Chat Input Visibility for Collaborators: Fixed an issue where the human escalation chat input in the inbox was not visible to collaborators with certain permissions.
May 16, 2024

Introducing Enhanced AI Capabilities and Flexible Training Options

πŸš€ New Features
  • Latest OpenAI GPT-4o Model: We have integrated the latest OpenAI GPT-4o model, the best in its class, now available to all users with access to GPT-4 series. This model offers a variety of context windows to suit diverse needs.
πŸ’‘ Improvements
  • Agent Retrain Approval: Previously, the AI Supervisor intents were automatically retrained with each update to the Agent name, description, or prompt. Based on user feedback, we've introduced a notification system that asks if bot makers would like to regenerate AI Supervisor intents on save, providing more control and flexibility.
To learn more about fine tuning intents, please visit: https://guide.gpt-trainer.com/fine-tuning-agent-intents
May 8, 2024

Seamless Chatbot Management and Expanded Notification Options

πŸš€ New Features
  • Transfer Chatbot Ownership: Users with the Pro plan can now transfer chatbot ownership to other users, provided that their subscription support's all of the chatbot's features.
πŸ’‘ Improvements
  • Email Notifications: You can now add multiple email addresses to send leads / variables data to multiple recipients.
May 2, 2024

Email Notifications and Real-Time Assistance

πŸš€ New Features
  • Email Notifications: Take your Sales / Customer Support workflow to the next level with the introduction of email notifications. GPT-trainer chatbots can now efficiently send email notifications containing your newly collected leads data. Find email notification settings in the Settings / Appearance section.
    • Daily Chat History: Stay updated with a daily email summary of the day's chat session history, ensuring you never miss important conversations.
    • Daily Collected Data History: Receive a daily email summary of all collected user data (or any variables you defined) across all chat sessions.
    • On-Demand Lead Forwarding: Get immediate access to newly collected leads' information or selected variables via email, enhancing your lead conversion.
  • Human Escalation Notifications: Our Human Escalation Agent now sends notifications through your browser when users request to chat with a real person. Whether you're actively using our app or taking a break, you'll receive notifications in any tab within your browser when someone requests human agent assistance. To set this up, refer to our dedicated guide here.
  • Unread Messages: Never overlook important messages again. We've added a badge to the Inbox sidebar tab to show the number of unread messages.
πŸ’‘ Improvements
  • Alerts UI: Improved UI for popup alerts.