This new release (BrainSoup 1.23.0) is a major update which introduces the following changes:

• New GuiAutomator tool: This tool allows your agents to interact with GUI applications. Your agents can now automate tasks in applications that don't have an API. This is a great way to automate repetitive tasks that involve interacting with graphical interfaces.
• New ImageEditor tool: This tool allows your agents to work with images. Your agents can now convert, resize, crop, rotate, flip, and apply filters to images and edit their EXIF metadata.
• New AudioVideoEditor tool: This tool allows your agents to work with audio and video files. Your agents can now convert, trim, merge, add audio or subtitles, extract audio or snapshots, apply effects, and more to audio and video files.
• New Camera tool: This tool allows your agents to see through the camera and answer questions about what they see.
• New Screenshot tool: This tool allows your agents to take screenshots of the screen and answer questions about what they see.
• New GeoCalculator tool: This tool allows your agents to convert between addresses and GPS coordinates and calculate distances between locations.
• New DateTimeCalculator tool: This tool allows your agents do date, calendar and time zone calculations.
• New FileArchiver tool: This tool allows your agents to compress and decompress files and folders in various formats.
• Improved Clock tool: This tool now has more features and takes less agent resources.
• Improved file handling: We have improved the way agents handle files, making all file-based tools more reliable.

This new release (BrainSoup 1.22.0) is a major update which introduces the following changes:

• New CsvRead and XlsxRead tool: Your agents can now extract data from CSV and Excel files using SQL queries, and can even join multiple files to consolidate data from different sources. These tools are very powerful and allow your agents to answer complex questions about your data. This is also a great way to connect your agents to your business data to automate tasks and provide insights.
• New SpeechToText tool: Your agents can now transcribe audio files to text. This uses OpenAI's Whisper model under the hood, and consumes very few credits. Additionally, if the file has already been transcribed during the document indexing process, the tool will not call the AI model, saving credits.
• New ImageToText tool: Your agents can now extract a detailed text description from images. This tool uses OpenAI's GPT-4o model under the hood. If the file has already been processed during the document indexing process, the tool will not call the AI model, saving credits.
• New Mistral models: We have added the open-mixtral-8x22b and codestral models to BrainSoup. open-mixtral-8x22b is very powerful and quite cost-effective, while codestral is specialized in code generation tasks, and very cheap to use.
• Cheaper pricing for Mistral models: We have updated the pricing for Mistral models. mistral-small and mistral-large are now 2 times cheaper!
• Cheaper image processing: We are now using the gpt-4o model instead of gpt-4-vision-preview as the default vision model. This model is twice as cheap, faster and more accurate.
• The 'Can see' and 'Can hear' permissions are now enabled by default for all new agents, as this has no negative impact on performance and usage costs.
• The 'FileSystem' context is now only injected into the agent's working memory if the sandboxed file system is not empty. This will prevent unnecessary context data from being sent to the agents, improving performance and usage costs.
• Optimized Doc-E's settings to improve its performance.
• Fixed a bug causing no information to be added to the message technical details when a tool usage fails. This will make debugging your tools easier.

This new release (BrainSoup 1.21.0) is a major update which introduces the following changes:

• PowerShell and CMD tools: Your agents can now use the 'PowerShell' and 'CMD' tools to run scripts and commands on the host machine. This very powerful feature allows agents to automate tasks, manage system configurations, use any command-line tool on the host machine, and perform a wide range of operations. Please note that these tools can be harmful for your system if used incorrectly, so use them with caution and only with well-instructed agents.
• Git Custom Tool: Your agents can now use the 'Git' tool to clone, pull, push, and manage Git repositories. This tool has been implemented as a Custom Tool, to demonstrate how easy it is to put any command-line tool at your agents' disposal.
• Improved small LLM performance: We have optimized the way the conversation history is handled by the LLMs, wich will improve the quality of the responses, especially for smaller LLMs. Please let us know if you notice any side effects.
• Faster agent response: We have optimized the RAG (Retrieval-Augmented Generation) process for better performance.
• Updated Doc-E's knowledge.
• Fixed a bug causing the message's technical details to only include the last tool used by the agent, when several tools were used.
• Fixed a bug causing the application to crash when trying to use an external language model that does not require an API key.
• Fixed the 'TextToImagePollinations' Custom Tool that was not working anymore since pollinations.ai changed its API.

This new release (BrainSoup 1.20.0) is a major update which introduces the following changes:

• New subscription plans: We have introduced new subscription plans to better suit your needs. You can now choose from a range of 3 plans starting at $5/month. You can switch between plans at any time by clicking on "Manage Subscription" in the BrainSoup app. If you subscribed before this change, you are now on the middle plan, which is exactly the same as your previous plan.
• Fixed bug causing some unwanted warning messages to be displayed when BrainSoup was started for the first time.

This new release (BrainSoup 1.19.0) is a major update which introduces the following changes:

• GPT-4o! We've added the latest OpenAI model to BrainSoup. This model is more powerful and more precise than its GPT-4 Turbo predecessors. What's more, it's 2 times faster... and half the price! As a result, the old GPT-4 Turbo model has been removed from BrainSoup and replaced with GPT-4o, all your agents will be automatically upgraded to the new model.
• Custom locations for Sandboxed File System: You can now choose a personalized location for a chat room's Sandboxed File System. This feature is especially useful for integrating with existing file structures or when working on shared projects.
• The old GPT-3.5 Turbo 16K model has been removed from BrainSoup, as the latest version of GPT-3.5 Turbo supports a 16K context window out of the box, for a lower price. All agents using the old model will be automatically updated.
• Fixed an issue preventing external LLMs from using tools, even when they have native function call support.
• Fixed a bug preventing you from having more than 3 active agents in the same chat room.
• Fixed a few other minor issues.

This new release (BrainSoup 1.18.1) is a major update which introduces the following changes:

• External LLMs! You can now integrate third-party LLM prodviders such as Hugging Face, Groq, or OpenRouter into BrainSoup. This feature allows you to access a broader range of AI capabilities beyond the built-in models, and take advantage of the latest advancements in AI without waiting for platform updates. You can learn more about this in our help article
• Doc-E on steroids: Doc-E now has full access to BrainSoup documentation, website, release notes and tutorials. This will allow Doc-E to provide more accurate and up-to-date information.
• Improved reliability: We have improved the validation/repair step applied to LLM JSON output. This will help make BrainSoup more robust and reliable.
• Fixed a bug causing unnecessary document repository scans.
• Fixed a bug causing image indexing to fail in some cases.

This new release (BrainSoup 1.17.0) is a major update which introduces the following changes:

• Enhanced user interface: In agent and chat room configuration screens, a short description is now displayed for each context subscription, event subscription and allowed tool.
• When an agent uses a tool, the full name (Tools + Function) is now displayed in the chat status and technical details of the message, for greater clarity.
• Updated Doc-E's knowledge.
• Fixed a problem in Ollama LLMs stop sequence management that sometimes caused agents to answer their own questions.

This new release (BrainSoup 1.16.1) is a major update which introduces the following changes:

• Local LLMs! You can now create agents using LLMs running on your own machine, for zero operating costs and improved confidentiality. The installation and management of these local models is left to the excellent free Ollama app, which you need to install separately. Once installed, all Ollama LLMs will be available in BrainSoup. You can also specify the URL of an Ollama server installed on another machine.
• Improved agent configuration window: The maximum token count of the selected LLM, working memory and agent responses are now displayed.

This new release (BrainSoup 1.15.0) is a major update which introduces the following changes:

• Multimodality for all LLMs! Support for audio and image files for richer interactions with your agents.
• New Embedding Vectors Model: We are now using the new OpenAI 'text-embedding-3-small' model, which is more efficient and 5 times cheaper than the previous one. Document indexing will now cost you almost nothing!
• Improved Document Indexing System: Smarter and 30 times faster indexing process, with support for text, audio and images.
• Support for New File Types: Including '.docx', '.wav', '.flac', '.mp3', '.mpga', '.m4a', '.ogg', '.jpg', '.jpeg', '.png', '.bmp', '.gif', enhancing data management.
• Multi-level Document Understanding: Your agents can now better interpret and interact with large content.
• Semantic Indexing of Sandboxed File System Files: Agents can now better understand the content of the files you share with them in the chat, espaecially when they are subcribed to the 'FileSystem' context.
• File Sharing with Agents: Easily share files with your agents by dragging and dropping them into the chat or the Sandboxed File System explorer.
• Confidential Conversations: You can now mark conversations as confidential to prevent their memorization by agents.
• Message Deletion: More effectively control the content of your discussions.

This new release (BrainSoup 1.14.0) is a major update which introduces the following changes:

• Mistral AI: All Mistral AI LLMs are now supported! From the very economical mistral-7b to the very powerful mistral-large, which rivals GPT-4. Our favorites are mistral-8x7b and mistral-medium, which offer the best cost/performance ratios for their use cases.
• Universal function calling: We've developed a system that allows any LLM to use tools, even if it wasn't designed to do so. When such an LLM needs to use a tool, it is able to delegate the function call to a superior LLM (GPT-4 at the moment, but we'll soon make this configurable). This system works for Mistral's LLMs, as well as for any future LLMs we add.
• Updated and optimized knowledge of Doc-E, to reduce it's operating costs.
• Fixed a problem that prevented message technical details from being copied to the clipboard.

This new release (BrainSoup 1.13.0) is a major update which introduces the following changes:

• Upgraded GPT-3.5-Turbo: BrainSoup now uses the latest version of GPT-3.5-Turbo, which has a 16K token limit, a better performance and an operating cost divided by 3! Consequently, the old GPT-3.5-Turbo-16K is now deprecated and will be removed in a future release.
• Acrylic window style: the BrainSoup's window can now have a beautiful transparency effect (can be disabled in the settings).
• Favorite chat rooms: you can now mark chat rooms as favorites, so that they appear at the top of the list.
• Multi-selection: you can now select several chat rooms at once to delete them or to mark them as favorites.
• Sorted chat room list: latest active chat rooms are now displayed at the top of the list.
• Better resource management: BrainSoup now consumes less memory and is much more responsive.
• Faster loading time: BrainSoup's startup sequence has been optimized and is now much faster.
• Background chat rooms now appear in bold in the chat room list.
• Several minor bugs fixed.

This new release (BrainSoup 1.12.1) is a major update which introduces the following changes:

• Enhanced Transparency with metrics: each message created by an agent in the chat now has a button for displaying technical information such as the token count, the cost, the memories used, etc.
• Revamped Tool Integration: overhaul of the system enabling agents to use tools. The new system is much more reliable, faster and ready to support new, non-OpenAI LLMs.
• Sleek main window title bar redesign.
• Fixed a bug that made it impossible to scroll with the mouse wheel on some computers.
• Fixed a bug that sometimes caused agents to answer their own questions.
• Fixed a bug that caused the application to crash when an agent tried to speak after its configuration had been modified.