Dify

?? Dify v0.12.0 is Out! ?? New Features 1?? File Variables in LLM Nodes Use File and Array File variables directly in the LLM Prompt editor within Workflows to seamlessly handle documents, images, and audio files (currently supporting gpt-4o-audio-preview). ?? Note: For LLMs without native file processing capabilities, you'll still need to use a Document Extractor node to convert files to text format. 2?? Document Processing for ChatBot & Agent Introducing document toggle for ChatBot and Agent applications! When your model supports document processing (txt, markdown, pdf, html, xlsx, etc.), enable this feature to let your application handle tasks based on user-uploaded files. ?? For Developers 1?? Python Environment Updates Default Python version upgraded to 3.12 with performance improvements and dependency updates (including Gevent). Now supporting Python 3.11-3.12. Python 3.10 tests removed for faster CI pipelines. 2?? App DSL Import API Changes - Deprecated: Legacy routes for DSL file/URL imports and Workflow override routes - Added: New /imports endpoints (affects users with extensive secondary development) For more improvements and bug fixes, read the full changelog: https://lnkd.in/gjJg6-4n

??? Watch this video and discover how to use Dify’s file upload feature in workflows to turn documents into AI-driven podcasts, delivering functionality similar to Google NotebookLM. ?? https://lnkd.in/gd2WXfnV

?? Dify v0.11.2 is here! Explore new models and tools contributed by our amazing community. ?? New Models - Gemini-exp-1114: Google’s latest top-tier model, ranked #1 on the lmarena.ai leaderboard for overall performance. - Added several Vision-Language Models (VLMs) from SiliconFlow, including qwen2-vl-72b and qwen2-vl-7b. JSON output format is now available. ?? New Tools - DuckDuckGo Video Search: Search for videos (e.g., from YouTube) directly in Dify Workflow and access transcripts. - FAL: Now integrates Wizper to transcribe audio files using the Whisper model. - Email: New tool for sending emails. For more improvements and bug fixes, read the full changelog: https://lnkd.in/gQBkbeUY

?? Join Dify at the AI Innovation Webinar! Our RAG Product Designer Yawen Zhao will reveal insights on building advanced RAG apps, including Dify’s upcoming Parent-child Retrieval feature, alongside experts from Amazon Web Services (AWS) and TiDB, powered by PingCAP. ?? Nov 28 ? 09:30 GMT ?? Online Register free: https://hubs.ly/Q02XQHYb0

?? Dify v0.11.1 comes with some news you might need to know: ?? Python Version Update We're upgrading to Python 3.12 in our next release. Both SaaS and Docker will run on 3.12 by default, with 3.11 compatibility maintained. Python 3.10 support will end. For source code deployment or custom development, please upgrade to 3.11/3.12. ?? New Features: - Tongyi Qianwen and Zhipu multimodal models now support video content. Create multimodal AI apps by enabling file uploads and using our new CogVideo tool to generate videos from text and images. - GitLab (build-in tool): Added merge request queries - OpenRouter: Added support for Claude 3.5 Haiku - App Icons: Now supports GIF, PNG, and WebP formats ? Read the full changelog: https://lnkd.in/gbqZ8i4A

???? The Iteration node can now handle arrays output from upstream nodes in parallel - meaning it can work on multiple items at the same time instead of doing them one after another. This makes everything run much faster. ??? We've also made it more reliable by implementing robust error handling. Now, if one task runs into trouble, it won't crash the whole process - everything else keeps running smoothly. ? Check out our blog on best practices for parallel and advanced iteration features: https://lnkd.in/gx_qbDGU

?? Dify v0.11.0 is out! ? Iteration Node Parallel Mode The Iteration Node now supports parallel processing with improved error handling, delivering a smoother, more efficient workflow experience. Docs: https://lnkd.in/g76XNxR6 Blog - Enhancing GPT-Researcher with Parallel and Advanced Iterative Features https://lnkd.in/gx_qbDGU ?? Dify Knowledge Base Retrieval API Dify Knowledge Base is now accessible as an independent service through its API. This feature empowers developers to query and retrieve chunks from their Dify knowledge bases. Cookbook: https://lnkd.in/gRfJqBby ?? Interactive Forms in Dialogues Create interactive forms right in your dialogues using Template Node, making it simple and intuitive for users to input information. Use case: https://lnkd.in/gdTZu3qw ?? New Integrations - Grok Beta (xAI) - 128K context length with function calling support - Claude 3.5 Haiku (Anthropic & AWS Bedrock) - Fast response time with SOTA coding capabilities - GPUStack models support - Vector Database: Couchbase, OceanBase, Lindorm See full changelog: https://lnkd.in/g5ymhg_5

?? Exciting times at #TechCrunchDisrupt! Our Head of Developer Relations, Chenhe Gu, took the stage to show how Dify empowers developers to turn #GenAI ideas into prototypes. We’re making AI application development easier than ever.

?? We’ve just hit 50k stars on GitHub! ?? A huge thanks to our incredible community for being part of this journey. We’re about to unveil something big in v1.0, something we’ve been working on day and night. It’s set to change the game, making Dify more open and accessible than ever. ?? For developers, adding new features to Dify will be more flexible—not just limited to PR submissions. ???? For AI app builders, Dify will soon allow easy connections to external apps and tools. Stay with us—we can’t wait to show you what’s next! ??

?? What an incredible day at the Moscone Center! We loved introducing Dify.AI to new friends and were thrilled to connect face-to-face with amazing Dify community members from around the world. So much energy and inspiration! We’ll be here for the rest of the event—visit us at booth K4!