Changelog

Boosting Agent Performance with New Models and Expanded Context Windows

• New Agent Models: We’re excited to introduce a powerful lineup of AI models for your agents! Now available: DeepSeek-R1, DeepSeek-V3, Gemini 1.5 Pro,Gemini 2.0 Flash, Claude 3.5 Haiku, Claude 3.7 Sonnet, GPT o1, and GPT o3 mini. These models come in various versions, supporting context windows ranging from 2k to an impressive 128k, giving your agents the flexibility to handle everything from quick chats to deep, context-rich conversations.

  We’ve also upgraded several existing models by increasing their context sizes.

  These additions and upgrades mean your AI agents are now more versatile and capable than ever. From snappy customer service replies to detailed conversational analysis, the new models and larger context windows unlock a world of possibilities for customization and efficiency.

Enhancing Chatbot Customization and Debugging with New Features

• Initial Messages Variables: Customize your chatbot’s initial messages with dynamic variables. Now, you can append user-specific data to tailor the chatbot’s first interaction based on different input parameters. "Hi {{{user_name}}}" instead of "Hi there!"`.
• Low Credit Notifications: Never run out of message credits unexpectedly! You can now enable low-credit notifications in your account settings, allowing GPT-Trainer to alert you when your balance is running low (applicable if you don't already have an AI API key set up).
• Widget Button Spacing Adjustments: Fine-tune the position of your chatbot widget! You now have the flexibility to adjust the bubble position to ensure optimal placement of the widget on your website.

• Extra Details in Debug Messages: Gain deeper insights into your chatbot’s decision-making process. We’ve enhanced debug messages to include RAG context, Google Search results, and error logs, helping you better understand the data sources behind responses and troubleshoot issues more efficiently.

• API Session Creation: Fixed an issue where the session creation API incorrectly required a mandatory JSON payload, which caused disruptions for some users. This has now been resolved.

Enhancing User Experience with New Features and Improvements

• UTM Tracking: You can now enable UTM tracking to capture and analyze the performance of your campaigns. Track user behavior across different marketing channels and campaigns to better understand where your traffic is coming from, and make data-driven decisions.
• SMS Integration: Link your phone number to your chatbot and start receiving messages instantly. This new SMS integration (in partnership with Ytel) allows you to configure GPT-trainer AI agents to auto-respond to incoming SMS messages.

• Export Tracking Data: We've made it easier to export your tracking data! You can now export a detailed list of all your tracking information in a convenient file.

New Dashboard Features, Chatbot Performance Enhancements, and Bug Fixes

• Upgraded Dashboard: ou can now track the visitors to your website with more detailed insights! The dashboard now displays the number of visitors and those who engaged in conversations. For Team plans and above, we’re adding even more granular data, including user location, browser, and OS information.

• Chat Widget Perfomance: We’ve made significant optimizations to the rendering process of our chatbot. The load size has been drastically reduced. Your widget should load faster.
• Redesigned Chatbot Templates: We’ve made the chatbot creation process much simpler! Our newly redesigned templates will preconfigure many settings for you based on your intended use case. You’ll get set up even more quickly than before!
• reference_source_link in API Data Upload: We’ve enabled the use of reference_source_link when uploading data sources via the API.

• Tags Not Copied When Duplicating Chatbots: We’ve fixed an issue where tags were not being copied when duplicating chatbots.
• Table Data Sources with Gemini: Fixed a bug where users could not use table data sources with Gemini, resolving compatibility issues.
• Human Escalation Load Issues: We’ve fixed an issue that caused loading problems with the Human Escalation feature.

New Features for Faster Chatbot Development and Easier Integration

• Duplicate Chatbots: You can now duplicate your existing chatbots! This feature is available in the chatbot card settings under the chatbot menu page. Please note, duplication is only possible if you are the owner of the chatbot.
• WordPress Plugin: We’ve successfully implemented a WordPress plugin for smoother integration. No more copying scripts manually—simply install the plugin, and your chatbot is ready to use directly in WordPress, hassle-free.

• More Accurate URL Citations: Chatbots now generate more precise and accurate citations when referencing external URLs in its training data.
• Less Sensitive Human Escalation: Human handoff triggers have been fine-tuned to reduce unintentional escalations.
• Improved Table Sources Processing: Chatbots now process and handle table data more effectively. This is part of our ongoing effort to optimize our RAG systems.

MAJOR Refactoring: Simplifying Our Product Experience

• Redesigned Settings/Appearance: The section has been transformed into Customizations, featuring a better style and easier access to all the tools that can help you set up your AI chat effectively.
• Background Agents Fully Integrated! Human Support Handover, Conversation Labeling, and User Data Collection have all been integrated natively into our multi-agent framework. You no longer have to create dedicated AI Agents for them. You can configure them inside the Customizations tab directly.
• User Data Collection & Variables: Variables used to belong under specific AI Agents. Now, we made variables global. This means that all variables are accessible to all AI Agents. Use Variables to collect data from conversation sessions naturally. Send the collected data to any external CRM or API via webhook!
• AI Agents Simplified: With most of the legacy AI Agent types now deeply integrated with our native framework, the only AI Agent type you need to configure is the user-facing AI Agent. You can use multiple agents of this type to create powerful chatbots with specialized agents trained to handle particular tasks.

Enhanced Source Management, AI Response Styles, and More!

• Source Tags: We’ve introduced the ability to create and assign tags to your sources, making it easier to manage and categorize them for improved organization. You can filter sources based on tags.
• AI Response Style Customization: Under the "Message Appearance" settings you can now choose between Streaming Response, where words appear as they’re generated like in ChatGPT, or Simultaneous Response, which shows a loading indicator and displays the full message once it’s ready. Simultaneous Response will let you "simulate" a human typing behind the chat, leading to a more personal interaction with your chatbot.

• Gemini Model Error Reporting: We’ve improved error reporting for all Gemini models.

• Creating and Editing Agents Permissions: We resolved an issue where collaborators encountered problems when creating new agents.
• User Identity Verification: We fixed a bug that was causing authorized user data to not pass correctly into chat sessions.
• Chats Timestamp: Resolved an issue where the correct time was not updating for certain chat integrations.

Exciting New Model Additions and Widget Customization Options

• Custom Widget Dimensions: You can now personalize the size of your chatbot widget by setting your own width and height using pixel dimensions. This allows for better alignment with your website’s design and improves user experience.
• Claude and Gemini Model Integration: We’re thrilled to introduce a variety of new models for your agents, including Claude Opus, Claude Haiku, Claude Sonnet, Gemini Flash, and Gemini Pro. These models are now available for testing, providing enhanced capabilities and flexibility for your chatbot interactions.

• Webhook Form Submission: We’ve fixed an issue where webhooks were not sending forms if the form appeared after a certain number of questions. This ensures that all your forms are submitted correctly, regardless of their position in the conversation flow.
• Google Drive and Meta Integrations: We resolved a problem where Google Drive and Meta integrations were being blocked by our security settings. These integrations should now function smoothly, allowing seamless access and communication across platforms.

Innovative Enhancements and Essential Updates for Your Chatbot

• LaTeX Rendering Support for Mathematical Expressions: We’ve added support for LaTeX within chatbot messages. The browser-based chat widget can now render and display LaTeX equations.

• Expanded Input Length Based on AI Model: Previously, chat input was limited to 2,000 characters, which could be restrictive for certain use cases. We’ve updated this limit so that it scales based on the smallest context you allow across all of your AI Agents.
• New GPT Models with Extended Context Windows:We’ve integrated new GPT models that support 32k and 64k token windows. These models provide your agents with the ability to access and process significantly more context. Gemini and Claude support are coming soon.
• Refined Chat Auto-Open Behavior: Our Auto-Open Chat feature has been refined for a more user-friendly experience. Instead of automatically opening the chat every time, the chat will now only auto-open if the user hasn’t explicitly closed it before.
• Enhanced App Security: We’ve made significant strides in improving our app’s security, focusing on preventing the insertion of malicious scripts or unwanted code that could disrupt functionality. Your chatbot’s security is our top priority, and these updates ensure a safer and more reliable user experience.
• Inbox Sessions Last Message Appearance: We’ve redesigned how messages appear in the Inbox chat list. Previously, users would see the first message their customers entered. Now, the latest message left in the chat is displayed.

• Collaborators Interaction in Facebook Apps:We’ve resolved an issue where collaborators were unable to send messages in chats connected to Facebook integrations, such as WhatsApp, Instagram, and Facebook Messenger. This fix ensures smooth and effective communication across all your integrated platforms, enabling your team to interact with customers seamlessly.

Innovative Enhancements and Essential Updates for Your Chatbot

• Image/Video Sources Training: Enhance your chatbot’s knowledge by training it using YouTube videos or images. By providing a YouTube video, the content will be extracted and used to train your chatbot, offering citations and sharing the video with your customers. Image uploads are now available from the Small Business+, Team, and Pro plans, allowing for up to 100 uploaded images per day. The AI will generate content based on these images and use them as references.
• Mark All Sessions as Read: Easily manage your chatbot sessions by marking all sessions as read with a single click in the Inbox settings.

• API Documentation: Our API documentation has been updated with more detailed and cleaner information, incorporating various improvements for better clarity and usability.
• Chatbot AutoOpen Option: This feature now includes options, allowing you to specify when the chatbot should open, whether on desktop, mobile, or both.
• Widget Form/Disclaimer: The user interface for forms has been enhanced to display asterisk icons on labels for required fields, along with smooth animations upon form submission for both form and disclaimer features.

• Widget Button Image: Resolved an issue causing incorrect image styles in Microsoft browsers.
• Widget Input: Fixed an issue that caused text to be cut off while typing in different languages (Korean, Chinese, Japanese, etc.) upon submission.
• API Chatbot Creation: Fixed the bug that caused issues when trying to create a chatbot using the API.