> For the complete documentation index, see [llms.txt](https://docs.lleverage.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.lleverage.ai/changelog/2024.md). # 2024 All the Lleverage features, updates, improvements and fixes released in 2024 {% updates format="full" %} {% update date="2024-12-31" %} ## December 2024 ### Enhanced Prompt Engineering Experience #### Prompt Node on Canvas We're excited to announce significant enhancements to Lleverage that improve workflow creation and prompt engineering capabilities, while expanding our model offerings. Here's what's new: * Inline Configuration: Configure your LLM settings, prompts, and variables directly within the workflow canvas * Compare Feature: Test different prompt variations side-by-side to identify the best performing version * Real-time Preview: See prompt results instantly as you make changes, making iteration faster and more efficient #### AI-Powered Prompt Generation We're introducing an intelligent prompt generation feature that helps you create production-ready prompts quickly: * Natural Language Instructions: Generate sophisticated prompts by simply describing your requirements in plain language * Best Practices Built-in: Generated prompts automatically incorporate prompt engineering best practices * Customizable Output: Fine-tune generated prompts to match your specific needs while maintaining their optimized structure ### New Model Support #### OpenAI GPT-4 Optimized (O1) Models We've added support for OpenAI's latest O1 model series: * Enhanced Performance: Access improved response quality and consistency * Optimized Speed: Benefit from faster inference times while maintaining high-quality outputs These updates significantly enhance your ability to build and optimize AI workflows in Lleverage. The new prompt engineering capabilities make it easier to create effective prompts, while the addition of O1 models provides more options for balancing performance, speed, and cost in your applications. We look forward to seeing how you leverage these new capabilities to build even more powerful and efficient AI solutions. *The Lleverage Team* Posted: December 6, 2025\ Version: 0.6.0 {% endupdate %} {% update date="2024-11-30" %} ## November 2024 ### Workflow Versioning We're excited to announce new capabilities in Lleverage that bring greater control over your workflows and expanded support for multi-modal data processing. Here's what's new: * Major and Minor Versions: Choose between Major and Minor version increments when deploying updates. Use Minor versions for small tweaks and Major versions for significant changes or new features. * Set Default Version: Designate a specific version as the default. The default version will use the main URL, while other versions will include their version number in the URL (e.g., `v2.1`). This makes it easy to control which version is live for users without breaking existing integrations. * Preview and Revert: Easily preview older versions to test prior functionality, and revert to a previous version if needed. Editing from any version automatically creates a new draft, so you can test changes without impacting the live workflow. These versioning features allow for more stable, flexible deployments, and ensure your workflows are always performing at their best. ### Multi-Modality: Image Processing Support We're expanding Lleverage's capabilities with multi-modal support—you can now process image files within your workflows alongside text and data. This means you can leverage models that analyze both visual and text-based data for a richer range of applications. * Supported Image Formats: Upload standard image types (e.g., JPEG, PNG) and process them using compatible AI models. This opens up new possibilities for tasks like image captioning, visual recognition, and text-image combination analysis. * Expanded Use Cases: Combine image data with other document types (PDF, Word, Excel, CSV) in your workflows to build powerful, multi-modal applications. This feature supports a variety of industries, including e-commerce (e.g., product image tagging), healthcare (e.g., image analysis), and more. These updates give you greater flexibility to build AI workflows that handle both text and image inputs, expanding the range of potential applications on Lleverage. This release empowers you to manage workflow changes with confidence through versioning, while unlocking new multi-modal capabilities that allow for more advanced, AI-driven insights across multiple data types. We look forward to seeing the innovative applications you build with these tools! *The Lleverage Team* Posted: November 15, 2025\ Version: 0.5.0 {% endupdate %} {% update date="2024-10-31" %} ## October 2024 ### New Models We're thrilled to introduce new AI models and powerful enhancements in this release, further expanding your capabilities in building AI-driven workflows. Here's what's new: #### Llama 3.1 and Llama 3.2 Instruct We've added support for Llama 3.1 and Llama 3.2 Instruct, two powerful language models designed for improved instruction-following behavior. These models are ideal for generating high-quality text, answering questions, or handling complex natural language tasks. * Llama 3.1 Instruct: Optimized for tasks requiring clarity, coherence, and context understanding, making it ideal for a wide range of use cases, from customer support to content generation. * Llama 3.2 Instruct: An even more refined version, offering faster response times and improved accuracy for complex instructions. #### Mixtral 8x22B Instruct and 8x7B Instruct In addition to Llama models, we're also introducing Mixtral 8x22B Instruct and Mixtral 8x7B Instruct models. These powerful models deliver high performance for AI tasks that require instruction-following capabilities, fine-tuned for diverse, complex problem-solving. * Mixtral 8x22B Instruct: Designed for handling large-scale tasks with high accuracy and contextual understanding. * Mixtral 8x7B Instruct: A smaller but highly efficient model suited for mid-scale operations with faster inference times. Both Mixtral models are perfect for workflows that need efficient, high-quality results across a wide variety of natural language tasks. ### Feature Update: Auto-Complete for Variables in Workflow Canvas We've introduced auto-complete for variables within the workflow canvas, making it easier and faster to build workflows. This new feature suggests available variables and inputs as you type, reducing errors and speeding up the workflow creation process. * Auto-Complete Suggestions: Get real-time suggestions for variables and inputs while building workflows. * Reduced Errors: Avoid manual typing mistakes by selecting variables from the auto-complete list. * Faster Workflow Building: Seamlessly reference and link variables across nodes with minimal effort, boosting your productivity. These new models and features are designed to enhance your AI workflows, making it easier to work with high-quality models and build more complex logic in a streamlined way. We're excited for you to try them out and look forward to your feedback! *The Lleverage Team* Posted: October 24, 2024\ Version: 0.4.0 {% endupdate %} {% update date="2024-09-30" %} ## September 2024 ### New Features We're excited to announce several new features and improvements in this release of Lleverage. These updates enhance your workflow management, data handling, and overall platform usability. #### Vector Stores We've introduced Vector Stores to help you manage and query vector embeddings directly within your workflows. Each client automatically gets a dedicated Postgres Vector Database, and you can now upload documents such as PDFs, CSVs, Word docs, and PowerPoints to generate vectors using OpenAI's embeddings model. * Automatic vectorization of uploaded documents (PDF, CSV, DOCX, PPTX) * Integration with Unstructured.io for document preprocessing, chunking, and summarization * Seamless integration with the Vector Store Similarity Search node, allowing for efficient semantic search and retrieval #### Data Sets Introducing Data Sets to easily manage tabular data within your workflows. You can now upload CSV files or manually enter data into structured tables that can be used for bulk testing, few-shot learning, or passing structured data into your workflows. * Upload tabular data through CSV or manual entry * Use data sets for bulk testing, batch processing, or as X-shot examples to guide model outputs * Easily integrate data sets into your workflows for automation and AI-driven tasks #### Projects With this release, we're introducing Projects as a new way to organize your work in Lleverage. Projects help you manage workflows, prompts, vector stores, data sets, connections, and more, all within a specific container that keeps resources and teams isolated. * Create and manage multiple projects within your organization * Invite team members to collaborate on specific projects with role-based access control * Projects come with their own connections, secrets, and environment variables to keep resources organized #### Horizontal Canvas Design We've revamped the workflow Canvas Design with a new horizontal layout! This update makes it easier to build, view, and manage complex workflows, especially when working with multiple nodes and chains of logic. * Enhanced horizontal canvas for a clearer, more intuitive view of workflows * Improved node linking and navigation, making it easier to visualize and manage large workflows * Optimized for smoother scrolling and node arrangement, helping you keep workflows organized as they scale These new features and improvements are designed to give you more control, better organization, and enhanced usability within the Lleverage platform. As always, we look forward to hearing your feedback and helping you leverage these tools to build more powerful AI-driven workflows. *The Lleverage Team* Posted: September 18, 2024\ Version: 0.3.0 {% endupdate %} {% update date="2024-08-31" %} ## August 2024 ### New Features #### Prompt Studio: Multi-variants * **Variant Creation**: Design and test multiple prompt versions simultaneously to identify the most effective approach * **Side-by-Side Comparison**: View outputs from different prompt variants in a comparative interface for easy evaluation * **Performance Metrics**: Track response quality, generation time, and token usage across variants * **Template Sharing**: Save successful variants as templates for future use or team collaboration * **Variant Branching**: Create new variants based on existing ones to refine and iterate your prompts methodically #### Workflow Builder **Document Parser Node** * **Multi-format Support**: Process documents in various formats including PDF, DOCX, TXT, and more * **Intelligent Extraction**: Automatically identify and extract key information from structured and unstructured documents * **Custom Parsing Rules**: Define specific rules for extracting data based on document layout or content patterns * **Metadata Handling**: Extract and utilize document metadata in your workflows * **OCR Integration**: Convert image-based text to machine-readable format for processing **Loop Node** * **Iterative Processing**: Process items sequentially through workflow segments * **Collection Handling**: Automatically iterate through arrays and collections of data * **Conditional Loops**: Set specific exit conditions to control loop execution * **Loop Counters**: Track iteration numbers and use them within your processing logic * **Batch Processing**: Handle large datasets efficiently by processing in manageable chunks **Parallel Node** * **Concurrent Execution**: Process multiple tasks simultaneously for improved efficiency * **Fork and Join**: Split workflow execution and merge results when all parallel tasks complete * **Load Balancing**: Distribute processing load across multiple parallel executions * **Error Handling**: Configure how errors in parallel branches should affect overall workflow execution * **Resource Management**: Control the maximum number of concurrent executions to optimize performance #### AI Helpers **Model Setting Advisor** * **Parameter Optimization**: Receive intelligent recommendations for temperature, top\_p, and other model parameters * **Use Case Matching**: Get parameter suggestions based on your specific workflow requirements * **Cost-Performance Balancing**: Find the optimal balance between model performance and token usage * **Interactive Guidance**: Answer questions about your needs to receive tailored setting recommendations * **Setting Explanations**: Learn about each parameter's impact on model behavior through clear explanations **Model Advisor** * **Task-Based Recommendations**: Get suggestions for the most suitable AI model based on your specific task * **Comparative Analysis**: View side-by-side comparisons of different models' strengths and weaknesses * **Budget Considerations**: Receive recommendations that account for your cost constraints * **Performance Metrics**: Make informed decisions based on speed, quality, and cost metrics for each model * **Specialized Capabilities**: Identify models with particular strengths in areas like code generation, creative writing, or factual responses **Variable Generator** * **Automatic Detection**: Identify potential variables in your prompts and workflows automatically * **Naming Conventions**: Generate consistent and descriptive variable names following best practices * **Type Inference**: Automatically detect and suggest appropriate data types for each variable * **Default Values**: Provide sensible default values based on variable context and usage * **Documentation**: Automatically generate descriptions for each variable to improve workflow clarity These enhancements to Lleverage provide more powerful tools for prompt experimentation, document processing, complex workflow logic, and AI-assisted configuration. The multi-variant testing in Prompt Studio enables faster optimization, while the new workflow nodes handle more sophisticated data processing scenarios. Our AI helpers streamline configuration decisions, making it easier to get the most out of your AI models and workflows. *The Lleverage Team* Posted: August 15, 2024\ Version: 0.2.0 {% endupdate %} {% update date="2024-07-31" %} ## July 2024 ### New Features #### Prompt Studio * Create and manage prompts with an intuitive interface designed for efficiency * Save and organize prompts in a centralized library for easy access * Test prompts directly within the studio to ensure optimal performance * Use variables within prompts to create dynamic, reusable templates * Share prompts across your organization to promote collaboration #### Workflow Builder * Design complex AI workflows through a visual, drag-and-drop interface * Connect multiple nodes to create sophisticated processing pipelines * Configure node settings with an easy-to-use properties panel * Test workflows at each stage to ensure proper functionality * Save and deploy workflows to automate repetitive tasks #### User Management * Implement role-based access control with User and Admin permission levels * Manage team members' access to specific features and resources * Generate invite links to seamlessly onboard new team members * Add users directly from the platform with email invitations * Monitor user activity and resource usage across your organization #### LLM Model Connections * Seamlessly integrate with leading language models: * OpenAI: Access GPT-3.5 and GPT-4 models for a wide range of text generation tasks * Gemini: Leverage Google's powerful models for advanced language understanding * Anthropic: Utilize Claude models for safe and helpful AI assistance * Configure model parameters directly within your workflows * Store API keys securely with enterprise-grade encryption * Manage rate limits and usage to optimize costs #### Database Connections * Integrate your workflows with popular database systems: * SQL: Connect to standard SQL databases for structured data storage and retrieval * PostgreSQL: Leverage advanced PostgreSQL features for complex data operations * Supabase: Utilize Supabase's developer-friendly platform for backend functionality * Execute custom queries directly from your workflows * Import and export data between databases and AI models * Store and retrieve AI-generated content for persistent applications *The Lleverage Team* Posted: July 10, 2024\ Version: 0.1.0 {% endupdate %} {% endupdates %} --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.lleverage.ai/changelog/2024.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.