Feedback

Hello Dify Community, I'm reaching out to suggest a new feature that addresses a significant challenge in user experience when working with Dify apps, particularly during code generation tasks. Is this request related to a challenge you are facing? Yes, the issue arises when generating large content that hits the maximum output token limit, causing the generation to abruptly stop, sometimes in the middle of a code block. Users can send "continue" to get a complete result but this results in an incomplete output which cannot be easily used or copied as intended. What is the feature you'd like to see? I propose two enhancements: A "continue generation" button that appears when content generation reaches the max token limit. This button should allow users to seamlessly continue generating content within the same message. - The button text should be internationalized and customizable within Dify Studio. An optional auto-continue feature that can be enabled to automatically extend content until a complete response is generated without hitting the token limit. How will this feature improve your workflow / experience? The implementation of these features would enable smoother and uninterrupted content creation, particularly beneficial for generating extensive outputs such as long-form articles or complete code blocks without manual intervention. Additional Context or Comments This functionality is available in ChatGPT and there are open-source tools for the "auto continue" feature, indicating a clear demand for such capabilities within generative AI applications like those built on Dify. While I am not able to contribute directly to developing this feature through coding or financial sponsorship, I believe its addition would greatly enhance user satisfaction and align with existing expectations set by other AI platforms. I look forward to hearing what others think about this idea and seeing how we can make using Dify apps an even better experience!

When doing changes to prompts, adding functions etc. during orchestrating of an AI assistant, it can be frustrating to forget to hit the "publish" button before moving away from the page. I suggest a implementation of a "Do you want to save" before navigation away. So users don't loose work they have done.

Hello Dify Community, I'd like to propose a feature that addresses a challenge I've encountered: selecting the correct model for Dify apps when faced with non-descriptive deployment names from model providers. Is this request related to a challenge you are facing? Yes, the challenge is the difficulty in identifying the correct model due to generic or non-descriptive deployment names that are carried over from model providers like Azure to Dify. What is the feature you'd like to see? I propose adding an optional field called displayName to the model setup form in Dify. This field would allow users to enter a more descriptive and user-friendly name for the model. If a displayName is provided, it should be displayed in the UI select inputs instead of the model name. If not provided, the model name would be used as a fallback. How will this feature improve your workflow / experience? This feature will significantly enhance the model selection process by allowing me—and potentially other users—to quickly identify the right model for our Dify apps. It also enables the creation of model aliases tailored to specific use cases, such as "coding assistant model" for coding-related tasks, which simplifies the selection process for non-experts. Additional Context or Comments The UI should prioritize displaying the displayName over the model name in all selection interfaces within Dify when both are available. This will ensure a user-friendly and efficient model selection experience. Can you help with this feature? I am open to assisting with testing and providing feedback to ensure the feature is implemented effectively and meets user needs. I believe this feature will be a valuable addition to Dify, enhancing usability and reducing confusion during model selection. I look forward to the community's thoughts and feedback on this suggestion.

Hello Dify Community, I'm thrilled to propose an extension to the Dify app's variables feature: the Advanced Variable Instructions . This feature will empower users to define multiple instruction blocks and combine them based on conditions, creating more dynamic and responsive interactions. Why Advanced Variable Instructions? Conditional Logic: Users can define blocks of instructions that are executed based on the presence or value of specific variables. This allows for more personalized and context-aware conversation flows. Complex Scenarios: With the ability to use conditions such as greater than, less than, or equal to, users can build advanced logic into their Dify apps, enabling them to handle a wide range of user inputs and scenarios. Enhanced User Experience: By tailoring the conversation flow based on the user's inputs, Dify apps can provide a more intuitive and personalized experience, leading to higher user engagement. Currently it requires prompt engineering to handle optional variables in the instructions because when the value is not provided, the LLM might not understand the instruction. Adding a instruction block only when a variable is set, makes it so much easier. Feature Enhancements: Expanded Variable Types: In addition to the existing variables like selects, short text inputs, and text paragraphs, we will introduce number and boolean variables to allow for easier definition of conditions. Dynamic Enabling of Features: Users can specify conditions for enabling or disabling advanced features such as RAG (Retrieval Augmented Generation) based on the values or presence of specific variables, providing more granular control over the functionality of their Dify apps. Incorporation into Expert Mode: To maintain simplicity for regular users, this extended functionality will be incorporated into the expert mode of the Dify app's prompt engineering section, providing a seamless transition for users who are ready to explore and implement more advanced logic. I look forward to hearing your thoughts and feedback on this revised feature suggestion! Let's build a more powerful and versatile Dify app together.

Hello Dify Community, I'm excited to propose a new feature for Dify: the Background Service Integration . This integration allows users to schedule and monitor background tasks within Dify, such as dataset syncing, health checks, backups, and webhooks processing, as well as extend it to AI-related tasks in the future. Key Features: Frontend Monitoring Page : Users can access a dedicated page within the Dify frontend to monitor ongoing and scheduled tasks. This page will also include the ability to schedule tasks, select supported handlers, set parameters, and filter tasks using tags. Job List and Detail Pages : The integration will provide a list page displaying all scheduled jobs and a detail page for each job, containing configuration details, previous and upcoming runs, logs, and health status to show how many jobs failed in the last few hours. Recurring Task Support : Users will be able to set recurring tasks, including those using crontab format, and see the next execution time for each recurring job, simplifying scheduling and planning. Integrated Core Tasks : The integration will support scheduling Dify's integrated core tasks and calling webhooks, making it a comprehensive solution for managing various types of background jobs. Default API for Schedulable Services : A default API will be provided for "schedulable services" with endpoints for scheduling events, fetching updates and logs, and performing health checks. Additionally, support for checking which features are supported by a service will be included to ensure seamless integration. Enhanced Features: Configurable Parameters : Users can configure the integration with specific parameters like API keys and host details, providing flexibility and customization for different use cases. Future AI Task Integration : While the initial focus is on core tasks, the integration will be designed to seamlessly extend to AI-related tasks in the future, such as scheduling model training and analysis jobs. Why Background Service Integration? Centralized Task Management : Users can manage all background tasks within Dify, streamlining scheduling, monitoring, and troubleshooting processes. Improved Visibility and Control : The frontend monitoring page and detailed job information offer improved visibility into ongoing tasks and enhance user control over task scheduling and execution

For example, when an application goes online and the privacy of users on the C end is involved, you need to disable the log collection function

I am looking for a way to execute multiple LLM end-to-end runs with different combinations of prompt templates, models, data, and hyperparameters in bulk. This feature, known as prompt sweeps, would greatly assist in debugging and optimizing LLM applications. I propose the addition of a prompt sweeps feature in the Dify playground. This feature would allow users to set up and execute multiple runs with different combinations of prompt templates, models, data, and hyperparameters. The results of these runs should be presented in a clear and organized manner for easy analysis and comparison. The ability to run multiple prompts with different models and datasets in a single operation would be a powerful tool for debugging and optimizing LLM applications. It would save time and provide a more efficient way to test and refine these applications.