Changelog
This section will be dedicated to new features and improvements made to Swiftask, including recent updates, bug fixes, as well as innovations and new developments that enhance the user experience.
📅 8 May 2025 🚀
✨ New reatures
📊 Diagram Download: You can now download AI-generated diagrams via the download icon in the chat.
📈 Mermaid Diagram Generation: Ask Swiftask to create Mermaid diagrams directly in your conversations.
🔐 Advanced Default Role Management for SSO: Workspace administrators can now set the default role assigned to users who join the workspace via SSO authentication (available on Team Growth and Team Starter plans).
👥 Customization of Onboarding Experience: Tailor the welcome of new members by automatically assigning them the appropriate permissions upon their first login.
🔧 Improvements
🖱️ Simplified Interface: Added a download button next to diagrams for quick exporting.
🔄 Better Visualization: Optimized display of diagrams in the chat interface.
How to use: Ask an agent to generate a Mermaid diagram, then click the download icon to save it.
🚀 Simplified Administration: Reduced time required to configure access for new users through automated role assignments.
🛡️ Enhanced Governance: Better control over access to resources and AI agents for new members joining via SSO.
🔄 Technical Changes:
⚙️ Updated Administration Interface: New section in the admin console for configuring default roles.
🔄 Improved SSO Compatibility: Optimized support for SAML attributes for better integration with enterprise identity solutions.
📅 4 May 2025 🚀
🧠 Integration of OpenAI o4-mini in AI Hub - Our platform now welcomes OpenAI's latest compact model, offering an excellent balance between performance and cost
🖼️ Native multimodal support - o4-mini can analyze images, diagrams, and text simultaneously, enabling richer and more contextual interactions
📊 Excellence in mathematics and programming - Exceptional performance on technical benchmarks, achieving 92.7% on AIME 2025 and 68.1% on SWE-bench Verified
🔍 Integrated web browsing - o4-mini can perform real-time searches to enrich its responses with current information
🎨 Image generation - Creation of images based on your textual descriptions, directly integrated into the conversation flow
🔄 Compatibility with existing agents - o4-mini can be selected as a model for your custom agents, enhancing their capabilities without modifying their configuration
📅 30 April 2025 🚀
🔧 Improvements
🔍 Enhanced skills search - Optimization of the search engine in the agent prompt section, making it easier to find relevant skills when configuring your AI agents.
🧩 Simplified navigation - More intuitive interface for adding skills to your agents, making the agent creation process smoother and more efficient.
🐛 Bug Fixes
🖱️ Drag and drop fix - Resolution of interface issues when dragging and dropping elements, ensuring a more stable and consistent user experience.
🔄 Improved stability - Fixed various minor issues related to drag and drop interactions across the platform.
📅 From 1 to 23 April 2025 🚀
✨ New features
🤖 New Semi-Sovereign Language Models
🇫🇷 Anthropic Claude 3.5 on Bedrock - Model hosted in France via AWS Bedrock, offering an excellent balance between performance and data sovereignty
🇫🇷 Mistral Large on Bedrock - Powerful version of the Mistral model, hosted in France via AWS Bedrock, for precise and contextual responses
🚀 New Advanced GPT Models
🧿 GPT-4.1 - Latest version of OpenAI's flagship model with enhanced reasoning capabilities (Team Starter and above plans)
📊 GPT-4.1 Mini - Compact version of GPT-4.1 optimized for quick responses while maintaining high quality (Starter and above plans)
📱 GPT-4.1 Nano - Ultra-lightweight version designed for applications requiring low latency and efficiency (Available on all plans, including Free)
➡️ OpenAI o4-mini - Available for Team Starter and above plans, offering optimized performance for team environments
🧠 New Gemini Models
💡 Gemini 2.5 - New generation of Google's multimodal model with improved contextual understanding
⚡ Gemini 2.5 Flash - Optimized version for instant responses with excellent performance
🧮 Gemini 2.5 Thinking - Specialized in deep reasoning and complex problem solving
🔬 Gemini 2.5 Experimental - Advanced version with cutting-edge experimental features
🔮 New Grok Model
🚀 Grok 3 Beta - Latest version of xAI's model with improved conversational and problem-solving capabilities, available for Starter and above plans
🧠 Long-term memory: New memory capability for both users and agents, enabling continuity in interactions and enhanced personalization
🔐 Secrets and variables management: Added functionality to securely store and manage sensitive information and environment variables
👥 Group governance: Introduction of a group structure with dedicated role management, offering precise access control (major feature for enterprises)
Each model has been carefully selected to offer different performance levels and price points, ensuring you can choose the right tool for your specific use case. The token coefficients have been adjusted to reflect the capabilities and processing power of each model.
For detailed information about pricing and specific features, please visit your Swiftask dashboard.
🔧 Improvements
🎨 Redesigned global interface: Lighter and more modern design for an optimized user experience
🛠️ Enhanced agent creation interface: Ability to directly select a skill when writing the prompt
🧭 Optimized agent management navigation: Better organization and more intuitive access to agent management features
👤 Advanced role management: New option to specify LLM and Embedding authorizations by role
📋 Improved projects/tasks interface: Redesign of the user interface for better visualization and management of projects and tasks
🏷️ Sovereignty Badges - All models are now clearly identified by badges indicating their level of data sovereignty
🌍 Sovereignty Filters - New option to filter models based on their level of data sovereignt
ℹ️ Detailed Information - Each model now has comprehensive information sheets on its technical characteristics and environmental impact
📝 Technical Notes
emi-sovereign models are hosted in France via AWS Bedrock, offering a good compromise between performance and compliance with European regulations
Workspace administrators can now limit access to certain models based on their organization's compliance requirements
Complete documentation on the characteristics of each model is available in our help center
We encourage you to explore these new models and provide feedback to help us continuously improve our platform.
📅 March 24, 2025 🚀
✨ New features
🏢 Added Pappers competency, enabling access to company information
📅 March 22, 2025 🚀
✨ New features
🔄 Task automation webhook allowing you to trigger an agent from your external application without necessarily using an API
⏸️ New ability to pause automations when needed
🐛 Bug fixes
🔗 Fixed issues related to agent sharing
📅 March 21, 2025 🚀
🐛 Bug fixes:
Fixed a bug with Google Drive, users can now browse through all folders in a Google Drive
📅 March 16, 2025 🚀
📅 March 15, 2025 🚀
📅 March 10, 2025 🚀
📅 March 7, 2025 🚀
Integration of Perplexity deep search in the Ai hub
Conduct much deeper searches with the new Perplexity Deep Search AI for even more relevant results.
Integration of GPT-4.5 in the AI hub and Agents
OpenAI's most advanced model is now available in the AI Hub Chat and Agent, offering:
- More natural language understanding
- Significantly expanded knowledge base
- Better ability to follow your intentions
- More developed "emotional intelligence"
- Significant reduction in hallucinations
- Enhanced performance for writing, programming, and problem-solving
📅 February 25, 2024 🚀
New AI for chat or LLM model for agents : Claude 3.7 Sonnet.
Its impressive versatility allows it to handle both simple calculations and the complete creation of websites, and it is immediately accessible via the Swiftask chat for natural conversations, as a basis for your customized AI agents without requiring programming, as well as among more than 80 integrated AI tools at your disposal.
📅 February 24, 2025 🚀
Importing Files from Dropbox and Google Drive: Swiftask connects directly to Google Drive and Dropbox to power your AI Agent with your documents. Instantly import your files, create customized agents, and interact in natural language. Easily analyze and query your data to get accurate answers.
📅 February 23, 2025 🚀
Bug Fixes: resolved issues related to uploading image files through the Chatbot widget, enhancing reliability and user experience.
New AI Perplexity with Deepseek R1 and Perplexity Pro with Deepseek R1 : experience enhanced reasoning, streamlined simplicity, and professional-grade performance with DeepSeek R1 technology.
New LLM models for Agents:
• xAI Grok 2: Elon Musk's model, direct and with a unique personality. • Google Gemini 2 Pro & Flash: combine power and speed for exceptional performance. • Claude 3.5 Haiku: one of the most powerful AIs, at a very low cost.
📅 February 22, 2025 🚀
Data source simplified interface: optimization of the interface for adding data sources, making the process more intuitive.
Improvements and Bug fixes in agent section: resolution of various issues concerning the sharing of agents, data sources, and skills, ensuring better stability and performance.
📅 February 17, 2025 🚀
Improvement of the agent skill "Retrieve data from external sources"
Possibility to add a data source in this skill, allows Swiftask AI agents to access company-specific data and facilitates the creation of more relevant and contextualized AI agents for each use case
📅 February 12, 2025 🚀
🐍 New agent skill: Python execution
Real-time Python code execution
Advanced programmatic operations for agents
📅 February 11, 2025 🚀
🔄 Onboarding optimization: simplified user invitation to Workspace
Now, when inviting a user, a temporary account is automatically generated with provisional credentials (email and password). This enhancement significantly streamlines the onboarding experience by:
Accelerating initial Workspace access
Reducing friction in the invitation process
Enabling immediate platform adoption
📅 February 8, 2025 🚀
📄 New agent skill: DOCX Export
Content export to DOCX format
Compatible with major word processors
Facilitates content sharing and reuse
📅 February 7, 2025 🚀
✨ New features
Conversation Starter to facilitate initial exchanges
Smoother initial interactions
Natural guide for starting conversations
Simplified drawer closing with outside click
New CSV Processor skill for data analysis
📅 February 6, 2025 🚀
🤖 Major Improvements
Redesigned agent creation interface
Optimized and more intuitive creation process
Simplified skill addition
Improved configuration interface navigation
New Widget version with optimized design
Advanced message management and usage limits
Support for file sending from widget
Privacy terms acceptance system
📅 February 1, 2025 🚀
New AI models
Integration of DeepSeek V3
Integration of DeepSeek R1
Last updated