Over the past month, Dify has rolled out several new features, further enhancing the product experience. In this summer special, we'll take you through Dify's latest updates and a sneak peek at some major upcoming features.
v1.7.0 | OAuth Support for Tools
Dify now supports OAuth authorization and Multi-credential management for tools.
OAuth for App Developers: when you need to integrate with services as Gmail, GitHub, and Notion that support OAuth, OAuth provides a more secure and convenient way to grant access compared to API keys. You can simply click to authorize the connection without having to manually find and copy a key.
Dify Cloud vs. Self-Hosted: Dify Cloud users can use a "Default" mode for popular plugins, as the Dify team has pre-configured OAuth applications. For self-hosted users, you'll need to register your own OAuth applications with third-party services and configure them in your Dify instance.
Multi-Account Management: Both versions allow you to link multiple accounts (like a work and personal email) and switch between them for different applications. A default credential mechanism ensures that if one is deleted, your application will automatically switch to another available one, ensuring stability.
For plugin developers, Dify has provided documentation on how to add OAuth support to your plugins.
v1.7.2 | Workflow Enhancements
Visual Debugging: The workflow editor now has a new relationships panel. You can select any node and hold the
Shiftkey to highlight its related nodes and connections, while fading out the rest. This makes it easier to trace data flow in complex workflows.
Quick Search and Navigation: Use
Cmd/Ctrl+Kto quickly search and jump to applications, knowledge bases, plugins, or workflow nodes. The system intelligently filters results based on your current page.
v1.8.0 | Smart Prompt & Code Fixes
Prompt and Code Optimization: Dify v1.8.0 introduces an assistant for LLM node prompt optimization and an auto-fix tool for Code nodes.
If a prompt isn't performing well, the optimization assistant can automatically generate an improved version based on your desired output.
If a Code node fails, the auto-fix feature will generate a corrected version of the code, saving you from having to debug in an external tool.
Both optimizations and fixes are saved as new versions, allowing you to easily compare and revert changes.
Workflow Default Values: The
Start nodein workflows now supports setting default values for all input fields. This means users can run a workflow immediately without entering any parameters, Greatly reduces the entry barrier for users.
In practical use, you can preset fixed cycles for a daily report generator or prefill common region and industry options for a customer service bot. Combined with the URL parameter prefill feature, this allows differentiated configurations across different channels. When upstream systems occasionally miss certain fields, default values serve as a fallback to ensure stable workflow execution.
Multi-Credential Management: You can now configure multiple API keys for a single model provider. This supports load balancing, so if one key reaches its limit or fails, Dify can automatically switch to a backup, ensuring stable service.
Accurate Usage Tracking: The Agent node now provides more accurate token usage statistics. Previously, all tokens were counted as output, leading to skewed cost estimates. Now, input and output tokens are calculated separately, allowing you to more precisely track costs. The usage data is also a structured output, which you can use in subsequent nodes to build automated cost-control features, like switching to a more economical model if token usage exceeds a certain threshold.
Performance and Stability: Dify's Workflow execution engine has been upgraded for significantly faster performance. We've made the database write operations asynchronous and non-blocking, which has drastically reduced the overall time it takes to run workflows. For complex workflows with parallel branches, the total execution time is now closer to the duration of the longest branch, rather than the sum of all branches. This upgrade also provides a more consistent streaming output and has greatly reduced timeout issues.
What's Next
Dify is planning to release a highly-anticipated feature: visualizing the RAG data processing pipeline. This will extend the workflow canvas experience to data handling, with a plug-in architecture that supports multi-modal data like text, images, and tables. This will enable businesses to create customized processes for converting various documents into high-quality context for LLMs.
