Pinned
Completed
Create Full-Stack app - Devv 2.0
We are developing a new feature of Devv, which uses Devv agent to generate full-stack app and deploy online. Here’s an example of using Devv agent to build a Mac OS simulator: We are planning to allow public beta by May. Leave a comment here if you are interested!
Bohan 4 months ago
High Priority
💡 Feature Request
Pinned
Completed
Create Full-Stack app - Devv 2.0
We are developing a new feature of Devv, which uses Devv agent to generate full-stack app and deploy online. Here’s an example of using Devv agent to build a Mac OS simulator: We are planning to allow public beta by May. Leave a comment here if you are interested!
Bohan 4 months ago
High Priority
💡 Feature Request
Pinned
Completed
Early bird price annual subscription devv pro
Annual payment subscription to devv ai pro, previously promised 12$ subscription price for early bird, again subscription payment shows that it is based on 15$.
Movit Randy 4 months ago
💡 Feature Request
Pinned
Completed
Early bird price annual subscription devv pro
Annual payment subscription to devv ai pro, previously promised 12$ subscription price for early bird, again subscription payment shows that it is based on 15$.
Movit Randy 4 months ago
💡 Feature Request
Suggestion for Improving the Code-Generation Quota Policy
Dear [Product / Support Team], I hope you are well. While using the service, I noticed that the monthly quota of 15 code-generation sessions is quickly exhausted. In my experience, more than half of these sessions are spent fixing bugs in previously generated code. Counting every debugging attempt against the same quota feels counter-productive and disproportionately raises the real cost of obtaining working code. To address this, I would like to propose two complementary ideas: Quota Deduction Based on Successful Deployment Only deduct a session from the quota when the generated code builds and deploys successfully. If the build fails, the attempt would not count, acknowledging that the output did not yet meet expectations. Introduce a Limited Retry & Rollback Mechanism • Allow a maximum of 3–5 repair attempts per original request. • Once the limit is reached, automatically roll back to the last version that deployed successfully. • The user can then make further changes from that stable baseline, preventing endless fixes from consuming the quota. These changes could protect platform resources while giving users confidence that each quota session truly advances them toward production-ready code. Thank you for considering this feedback. I am happy to provide additional details if helpful. Best regards, MarkHoo markhoo@foxmail.com
dhjtyjhtdr About 2 hours ago
💡 Feature Request
Suggestion for Improving the Code-Generation Quota Policy
Dear [Product / Support Team], I hope you are well. While using the service, I noticed that the monthly quota of 15 code-generation sessions is quickly exhausted. In my experience, more than half of these sessions are spent fixing bugs in previously generated code. Counting every debugging attempt against the same quota feels counter-productive and disproportionately raises the real cost of obtaining working code. To address this, I would like to propose two complementary ideas: Quota Deduction Based on Successful Deployment Only deduct a session from the quota when the generated code builds and deploys successfully. If the build fails, the attempt would not count, acknowledging that the output did not yet meet expectations. Introduce a Limited Retry & Rollback Mechanism • Allow a maximum of 3–5 repair attempts per original request. • Once the limit is reached, automatically roll back to the last version that deployed successfully. • The user can then make further changes from that stable baseline, preventing endless fixes from consuming the quota. These changes could protect platform resources while giving users confidence that each quota session truly advances them toward production-ready code. Thank you for considering this feedback. I am happy to provide additional details if helpful. Best regards, MarkHoo markhoo@foxmail.com
dhjtyjhtdr About 2 hours ago
💡 Feature Request
Memory issues of the agent
Hello, I’ve been using Devv for a while now — thank you for this tool, it’s really good. However, I’ve recently noticed a significant drop in the agent’s context memory. Previously, I could have long conversations without needing to repeat myself. But over the past few days, this has changed — now the agent seems to lose context after just 1 or 2 messages. This is a noticeable difference compared to how it used to perform. Is this expected? Thank you!
Alessandro About 1 month ago
🐛 Bug Report
Memory issues of the agent
Hello, I’ve been using Devv for a while now — thank you for this tool, it’s really good. However, I’ve recently noticed a significant drop in the agent’s context memory. Previously, I could have long conversations without needing to repeat myself. But over the past few days, this has changed — now the agent seems to lose context after just 1 or 2 messages. This is a noticeable difference compared to how it used to perform. Is this expected? Thank you!
Alessandro About 1 month ago
🐛 Bug Report
problem with extension and icon
I have a problem, I install the extension, but it won't appear on the side like every ai model when it is added to visual studio code, you set it that way or I have a problem
niko1984sd 15 days ago
💡 Feature Request
problem with extension and icon
I have a problem, I install the extension, but it won't appear on the side like every ai model when it is added to visual studio code, you set it that way or I have a problem
niko1984sd 15 days ago
💡 Feature Request
输入的时候怎么换行？輸入的時候怎麽換行？How to wrap when inputting?
Chrome 浏览器，Mac OS，无论输入 Enter 还是 Command+Enter，Option+Enter，Ctrl+Enter，Shift+Enter 都无法换行，都是直接提交，导致我每次要提交长文本排版都要先在本地文档编辑好之后，再复制到输入框。这么基础的功能都没有吗？ Chrome 瀏覽器，Mac OS，無論輸入 Enter 還是 Command+Enter，Option+Enter，Ctrl+Enter，Shift+Enter 都無法換行，都是直接提交，導致我每次要提交長文本排版都要先在本地文檔編輯好之後，再復製到輸入框。這麽基礎的功能都沒有嗎？ Chrome browser, Mac OS, no matter I type Enter or Command+Enter, Option+Enter, Ctrl+Enter, Shift+Enter, I cannot wrap the text, it just submits directly, which means I have to edit it in the local document first and then copy it to the input box every time I want to submit a long text layout. Don't they have such a basic function?
Ricky Song 16 days ago
💡 Feature Request
输入的时候怎么换行？輸入的時候怎麽換行？How to wrap when inputting?
Chrome 浏览器，Mac OS，无论输入 Enter 还是 Command+Enter，Option+Enter，Ctrl+Enter，Shift+Enter 都无法换行，都是直接提交，导致我每次要提交长文本排版都要先在本地文档编辑好之后，再复制到输入框。这么基础的功能都没有吗？ Chrome 瀏覽器，Mac OS，無論輸入 Enter 還是 Command+Enter，Option+Enter，Ctrl+Enter，Shift+Enter 都無法換行，都是直接提交，導致我每次要提交長文本排版都要先在本地文檔編輯好之後，再復製到輸入框。這麽基礎的功能都沒有嗎？ Chrome browser, Mac OS, no matter I type Enter or Command+Enter, Option+Enter, Ctrl+Enter, Shift+Enter, I cannot wrap the text, it just submits directly, which means I have to edit it in the local document first and then copy it to the input box every time I want to submit a long text layout. Don't they have such a basic function?
Ricky Song 16 days ago
💡 Feature Request
I recently used an AI-powered platform to create a website, and I have mixed feelings about the experience.
Certainly! Here is a review for an AI-powered website creation platform: I recently used an AI-powered platform to create a website, and I have mixed feelings about the experience. Pros: Design Capabilities: The platform excels in creating visually appealing designs. The AI seems to have a good understanding of modern design principles, which results in attractive and user-friendly layouts. Component Knowledge: The AI demonstrates a strong grasp of various web components and how to integrate them effectively. This makes it easier to create complex and functional websites. Ease of Use: The interface is intuitive, and the AI provides helpful suggestions, making the website creation process smoother for users who may not have extensive technical knowledge. Cons: Overcomplication: One of the main issues I encountered was the AI's tendency to overcomplicate things. I specifically requested a simple HTML + CSS template with a well-designed homepage. However, the AI kept generating almost full-fledged applications with numerous components and React functionalities that I did not need. Lack of Customization Control: While the AI is great at creating complex designs, it sometimes ignores specific user requirements. It would be beneficial to have more control over the level of complexity and the type of output generated. Suggestions for Improvement: Simpler Output Options: It would be helpful if the platform offered an option to generate simpler, more straightforward templates without unnecessary components. This would cater to users who need basic but well-designed websites. Better Adherence to User Specifications: The AI should more closely follow the user's specifications. If a user requests a simple HTML + CSS template, the AI should prioritize delivering that rather than adding extra functionalities. Customization Settings: Introducing settings that allow users to specify the level of complexity they need could enhance the user experience. This way, users can choose between basic templates and more complex applications based on their requirements. Overall, the platform has impressive design capabilities and a good understanding of web components. However, it needs to improve in adhering to user specifications and providing simpler output options for those who do not require full-fledged applications. Thx dev
Dune Stat 17 days ago
📥 Feedback
I recently used an AI-powered platform to create a website, and I have mixed feelings about the experience.
Certainly! Here is a review for an AI-powered website creation platform: I recently used an AI-powered platform to create a website, and I have mixed feelings about the experience. Pros: Design Capabilities: The platform excels in creating visually appealing designs. The AI seems to have a good understanding of modern design principles, which results in attractive and user-friendly layouts. Component Knowledge: The AI demonstrates a strong grasp of various web components and how to integrate them effectively. This makes it easier to create complex and functional websites. Ease of Use: The interface is intuitive, and the AI provides helpful suggestions, making the website creation process smoother for users who may not have extensive technical knowledge. Cons: Overcomplication: One of the main issues I encountered was the AI's tendency to overcomplicate things. I specifically requested a simple HTML + CSS template with a well-designed homepage. However, the AI kept generating almost full-fledged applications with numerous components and React functionalities that I did not need. Lack of Customization Control: While the AI is great at creating complex designs, it sometimes ignores specific user requirements. It would be beneficial to have more control over the level of complexity and the type of output generated. Suggestions for Improvement: Simpler Output Options: It would be helpful if the platform offered an option to generate simpler, more straightforward templates without unnecessary components. This would cater to users who need basic but well-designed websites. Better Adherence to User Specifications: The AI should more closely follow the user's specifications. If a user requests a simple HTML + CSS template, the AI should prioritize delivering that rather than adding extra functionalities. Customization Settings: Introducing settings that allow users to specify the level of complexity they need could enhance the user experience. This way, users can choose between basic templates and more complex applications based on their requirements. Overall, the platform has impressive design capabilities and a good understanding of web components. However, it needs to improve in adhering to user specifications and providing simpler output options for those who do not require full-fledged applications. Thx dev
Dune Stat 17 days ago
📥 Feedback
Completed
Claude 4 is available now!
I would like to see new models on the platform soon. Claude 3.5 sonnet is still an amazing coding assistant, please keep it for a while for comparison.
AirstaN About 2 months ago
💡 Feature Request
Completed
Claude 4 is available now!
I would like to see new models on the platform soon. Claude 3.5 sonnet is still an amazing coding assistant, please keep it for a while for comparison.
AirstaN About 2 months ago
💡 Feature Request
Requires a method or documentation for using the full version of the expansion plugin
devv replied I'm using the 1.0.0-mock extension, and I'd like to use the full version of the extension to try out the ability to modify local code directly. Please send me the way to use it
jcdl369 29 days ago
💡 Feature Request
Requires a method or documentation for using the full version of the expansion plugin
devv replied I'm using the 1.0.0-mock extension, and I'd like to use the full version of the extension to try out the ability to modify local code directly. Please send me the way to use it
jcdl369 29 days ago
💡 Feature Request
UI Enhancement Suggestions to Improve User Experience
Hello, I would like to suggest several UI improvements that would enhance the user experience: For thinking models, please add a toggle for enabling/disabling the thinking mode (for models where this feature is available). The colour scheme became too simple and lacks necessary contrast. It is now difficult to perceive code in light colours only, as there used to be a contrast with black. This reduction in contrast impacts readability during coding sessions. The edit mode now displays a super small window with scroll functionality, which makes it very inconvenient to copy text from there and edit effectively. A larger editing area would greatly improve usability.
AirstaN About 1 month ago
💡 Feature Request
UI Enhancement Suggestions to Improve User Experience
Hello, I would like to suggest several UI improvements that would enhance the user experience: For thinking models, please add a toggle for enabling/disabling the thinking mode (for models where this feature is available). The colour scheme became too simple and lacks necessary contrast. It is now difficult to perceive code in light colours only, as there used to be a contrast with black. This reduction in contrast impacts readability during coding sessions. The edit mode now displays a super small window with scroll functionality, which makes it very inconvenient to copy text from there and edit effectively. A larger editing area would greatly improve usability.
AirstaN About 1 month ago
💡 Feature Request
中文问题，中文回答，但是偶残渣俄文。
能够打通小周天的人有多少。 这是我的提问。 下面是回答： 关于有多少人能够真正“打通小周天”，这是一个很难给出确切数字的问题，原因如下： 定义模糊：什么是“打通小周天”本身就存在不同的理解。不同流派、不同传承对“打通”的标准和体验描述可能不一致。有些人可能只是体验到一些气感，而另一些人则可能达到更深层次的能量循环。 主观体验：小周天是一种主观的身体体验，很难用客观标准来衡量。即使是经验丰富的老师，也难以准确判断学生的真实情况。 隐秘性：传统上，气功修炼被认为是比较私人的事情， практикующие обычно не афишируют свои достижения. 因此，即使有人 действительно 打通了小周天，也未必会公开。 缺乏统计：目前没有任何机构或研究对 практикующие 气功的人数以及他们 достижение 小周天的情况进行过大规模的统计。 尽管无法给出确切数字，但可以肯定的是，真正能够“打通小周天”的人数相对较少。 以下是一些可能影响“打通”成功率的因素： 正确的指导：明师的指导至关重要。没有正确的指导， практикующие 可能会走弯路，甚至出现偏差。 持之以恒的练习：小周天 не достигается 一蹴而就，需要长期坚持练习，积累足够的能量。 身心状态： практикующие 需要保持身心放松、平静，才能更好地感知和引导能量的流动。 个体差异：每个人的身体状况、 энергетическая чувствительность 和领悟能力都不同，这也会影响 их прогресс. 总而言之，虽然“打通小周天”被认为是气功修炼的一个重要里程碑，但它 не является 容易实现的目标。需要正确的指导、 длительные усилия 和良好的身心状态。 之前就偶尔有俄文，但是只有一两个单词，这次这个答案有很多俄文
fsdgfsdgfsdfds About 1 month ago
💡 Feature Request
中文问题，中文回答，但是偶残渣俄文。
能够打通小周天的人有多少。 这是我的提问。 下面是回答： 关于有多少人能够真正“打通小周天”，这是一个很难给出确切数字的问题，原因如下： 定义模糊：什么是“打通小周天”本身就存在不同的理解。不同流派、不同传承对“打通”的标准和体验描述可能不一致。有些人可能只是体验到一些气感，而另一些人则可能达到更深层次的能量循环。 主观体验：小周天是一种主观的身体体验，很难用客观标准来衡量。即使是经验丰富的老师，也难以准确判断学生的真实情况。 隐秘性：传统上，气功修炼被认为是比较私人的事情， практикующие обычно не афишируют свои достижения. 因此，即使有人 действительно 打通了小周天，也未必会公开。 缺乏统计：目前没有任何机构或研究对 практикующие 气功的人数以及他们 достижение 小周天的情况进行过大规模的统计。 尽管无法给出确切数字，但可以肯定的是，真正能够“打通小周天”的人数相对较少。 以下是一些可能影响“打通”成功率的因素： 正确的指导：明师的指导至关重要。没有正确的指导， практикующие 可能会走弯路，甚至出现偏差。 持之以恒的练习：小周天 не достигается 一蹴而就，需要长期坚持练习，积累足够的能量。 身心状态： практикующие 需要保持身心放松、平静，才能更好地感知和引导能量的流动。 个体差异：每个人的身体状况、 энергетическая чувствительность 和领悟能力都不同，这也会影响 их прогресс. 总而言之，虽然“打通小周天”被认为是气功修炼的一个重要里程碑，但它 не является 容易实现的目标。需要正确的指导、 длительные усилия 和良好的身心状态。 之前就偶尔有俄文，但是只有一两个单词，这次这个答案有很多俄文
fsdgfsdgfsdfds About 1 month ago
💡 Feature Request
Need improvement can upload code
I pasted the code and it says keyword limit, I find it really annoying and makes my interaction with the AI difficult to cooperate
Taz Richard About 2 months ago
💡 Feature Request
Need improvement can upload code
I pasted the code and it says keyword limit, I find it really annoying and makes my interaction with the AI difficult to cooperate
Taz Richard About 2 months ago
💡 Feature Request
Can large language models support longer context in a single conversation? 36k is far from sufficient for code editing; and sometimes even before reaching 36k, there's a red box prompt preventing further input, which is quite frustrating.
muxb kongming 2 months ago
💡 Feature Request
Can large language models support longer context in a single conversation? 36k is far from sufficient for code editing; and sometimes even before reaching 36k, there's a red box prompt preventing further input, which is quite frustrating.
muxb kongming 2 months ago
💡 Feature Request
Data retention
Hi, could you tell me the data retention policy? Is it possible for me to opt out? Thanks.
Zachary Tang About 2 months ago
💡 Feature Request
Data retention
Hi, could you tell me the data retention policy? Is it possible for me to opt out? Thanks.
Zachary Tang About 2 months ago
💡 Feature Request