External reviews are not included in the AWS star rating for the product.
My main use case for Dify is to explore its capability and understand what it can do, as I primarily use it for uploading documents and obtaining answers.
A specific example of a document I uploaded to Dify is archive AI papers, and I wanted to ask about reinforcement learning.
The best features Dify offers, in my opinion, are trust and the ability to upload a large volume of documents.
Since I have used Dify personally, I have noticed ways it has made my workflow and research easier, saving me time and helping me find answers more quickly.
The document upload process works when I uploaded a bunch of PDFs, but ingestion is slower than I expected; I had 20,000 PDFs, and it could only process a few hundred in a day.
Dify could be improved, as mass ingestion could be handled better if you have thousands of documents to upload.
In my experience, Dify is stable, as I have not encountered any crashes or bugs.
Dify does not appear to handle larger workloads well regarding its scalability.
Before trying Dify, I tried LangChain and similar solutions; I switched because they were too manual.
My experience with pricing, setup cost, and licensing is that I only tried the free version.
I have seen a return on investment so far, as Dify worked for me, but I do not have a metric to determine how much time it saved.
I would add that Dify appears to be a very well-built enterprise-class product with good visibility.
I rate Dify overall as eight out of ten.
I chose eight out of ten because if it were faster and provided better answers, it would bring it closer to a ten for me.
My advice to others looking into using Dify is to be very clear on what purpose you intend to use it for.
I do not have any additional thoughts about Dify before we wrap up.
I tried Dify to build a specialized linear template architecture, which takes unstructured email data as input, identifies the top three trending narratives like OpenAI news or Anthropic news, and instantly transforms it into three distinct social-ready drafts that can be directly posted on LinkedIn.
I was planning to set up complete automation with Dify, but due to a lack of time, I tried the system with user input instead.
Since I was using Dify for the first time, I started by providing manual input for real-time analytics, but my main goal was to automate the process so that whenever an email from company X drops, Dify would automatically take the user input and generate LinkedIn posts.
I have not yet integrated Dify with any third-party applications, but I am considering integrating it with LinkedIn so that when I receive tech news emails, it can automatically generate and post them.
I compared Dify with both N8N and Zapier, and found that while Zapier lacks a deep prompt engineering environment and template notes, Dify allows for easy access to template notes to control the nuances of AI creative output. Additionally, N8N can become overly complex for content teams, while Dify provides a more specialized LLM narrative canvas, making it far better than both N8N and Zapier. I also received extra free credits compared to them.
In my work, I uploaded three news items to test Dify's ability to handle massive datasets, and despite the potential complexities of the connections between the news, Dify created accurate posts based on those inputs.
Before discovering Dify, I found Zapier and N8N too complex, and I had created a custom GPT that only took one news item at a time to generate LinkedIn posts. However, Dify minimizes human error by allowing me to generate approximately four to five posts in one go rather than inputting data for each news individually.
Dify can be improved by adding features such as an end loop or exit loop capability, similar to options available in N8N and Zapier, to make workflow completion easier without needing to select additional outputs and understand complex steps, which can be time-consuming.
I would appreciate an end loop button in the next step section of template 2 or iteration 2.
I think Dify can easily adapt, though I see potential for improvement, such as integrating features from Gemini to simplify workflows without needing to copy or edit settings manually.
Specifically, I want Dify to integrate with Gemini, as it would allow for effortless prompting without additional steps of copying or pasting between platforms.
I have been using various AI tools for the past four months, so I got into the field of artificial intelligence four months ago and try to use at least one new tool daily, which led me to try Dify after reading about it in the news.
I have not experienced outages or latency issues, except for the one problem I faced with the end node.
I have not contacted Dify's technical support because I have utilized the interactive mode in Gemini for issue resolution.
I solely relied on Gemini for assistance and did not use any official documentation from Dify. I already understood how to use user input, the LLM model, and templates without needing guides.
Before discovering Dify, I found Zapier and N8N too complex, and I had created a custom GPT that only took one news item at a time to generate LinkedIn posts. However, Dify minimizes human error by allowing me to generate approximately four to five posts in one go rather than inputting data for each news individually.
I performed all setups for Dify directly from the web.
Solving the end node issue took me 15 to 20 minutes, but once I resolved it, I received the perfect output.
I compared Dify with both N8N and Zapier, and found that while Zapier lacks a deep prompt engineering environment and template notes, Dify allows for easy access to template notes to control the nuances of AI creative output. Additionally, N8N can become overly complex for content teams, while Dify provides a more specialized LLM narrative canvas, making it far better than both N8N and Zapier. I also received extra free credits compared to them.
As a student, I would consider automating or selling with Dify despite its higher pricing compared to N8N and Zapier. However, if I had a team with more experienced members, I would likely opt for N8N or Zapier due to their lower pricing.
I did not use any metrics to determine the success of automated report generation in Dify.
I find Dify's pricing quite flexible compared to Zapier and N8N. As a daily AI user, the number of credits I get for free is important to determine how the tool can perform for me, and in this case, Dify is better than both N8N and Zapier.
The reliability of Dify's output depends on the LLM model I am using, and I find its architecture works perfectly for my needs.
Considering the pricing, I would rate Dify a seven on a scale from one to ten, but for usability, use case, and feasibility, I would give it a ten, as it surpasses both Zapier and N8N. My overall review rating for Dify is nine out of ten.
I am using Dify to automate some of my tasks, for example, sending emails to my clients, reaching out to new customers, and creating end-to-end automation for my clients.
Recently, I helped one of my clients from Denmark automate his end-to-end BI task using Dify, where I connected some EU-hosted inference providers and we built an end-to-end BI pipeline using Power BI modeling MCPP server. Now he does not need to check the dashboard every day; instead, he gets notifications on Slack, and AI completes all the data analysis and sends alerts in Slack.
Since we have a very small team, when we analyze our data, we cannot hire a proper BI and data team. With Dify, we have built automation that acts as a data analyst.
The first feature Dify offers, in my experience, is the ability to bring your own model and connect to different tools, making it very simple to build and get started. That is why I prefer Dify compared to other automation tools such as n8n or Flowise or some other providers.
Compared to other tools, for example, n8n, which promotes that it is very easy for non-technical people to start using AI and build their own automation, I have found that when you start with n8n, you still require technical skills. On the other hand, Dify is very handy; once you log in, you can use some pre-built templates to get started, and that is what I truly appreciate about Dify.
The interesting projects and the flexibility I appreciate about Dify are that you can bring your own inference provider and model provider. I am in the EU, and that is the biggest problem we face: we always have to use AI while complying with GDPR, General Data Protection Regulation. It is very easy to bring your own LLM provider and build automation in Dify, which is what I appreciate about it.
Dify stands out to me because it is compliant with GDPR, and it is 100% compliant with GDPR rules. I have not found any documents or documentation on other providers such as n8n and some other automation tools. Dify clearly explains how they are processing the data, which is why I appreciate it and why I am rating Dify a ten.
With the AI-powered data analysis we built, we achieve cost savings since we do not need to hire a person specialized in data analysis. It also results in time saving and cost saving.
Adding more pre-built advanced templates, progressing from basic to advanced templates, will help new users onboard faster. People who are already using Dify can quickly pick the pre-built templates as well.
For EU customers, adding more documentation about how Dify processes the data when starting to use Dify would be really beneficial for companies in Europe to get started with Dify.
I have been using Dify for almost a year.
Dify is stable in my experience.
Dify's scalability is very high, so you do not need to think too much about scalability.
Customer support is also great, which is one of the things I appreciate about Dify; whenever you have questions, you get an instant answer.
I did not switch to any different solution apart from Dify; I have tried others and then I came back to Dify.
When automating something, you are indirectly converting manual tasks into automation, which means you do not need to repeat those tasks manually. This results in cost saving and time saving, as whenever you save some time, that is equal to cost saving.
The experience with pricing is that it is quite reasonable, and the setup cost depends on our usage, so I do not have any complaints regarding this.
I evaluated n8n before choosing Dify.
I would definitely advise whoever is going to start with Dify to begin with some basic automation and then progress to more advanced automation. I am rating Dify a ten out of ten. I am rating customer support a nine out of ten.
My main use case for Dify is automation like HR automation, CRM automation, and ERP level management, which have been integrated into SaaS applications.
I have used Dify for HR automation and specific examples related to that can be provided.
Dify offers several best features including being very seamless, open source, having multiple plugins and a marketplace to use, and allowing integration of multiple AI models with full flexibility.
Dify's flexibility helps with automations or integrations since it is very seamless and has a very simple UI so anyone can use it, without anything too complex.
Dify stands out because it supports all the major AI platforms and plugins, and it has a marketplace where there are multiple things that can be integrated, which are official partners as well.
Dify has positively impacted the organization because accuracy has been improved, and the time and complexity in flows that were manual are now automated, from HR automation to ERP level transactions, including subscription management in the SaaS application, monitoring, and analytics.
There are no exact numbers available, but the time has drastically been reduced and the performance has improved since Dify was introduced in the system.
No features are needed that Dify doesn't have as it has all the features required.
Everything is working well with Dify. The only improvement would be if Dify provided an SMTP server that could be connected to automate Dify workflow management, as that would be a great option.
I have been using Dify for the past one year.
Dify is stable and scalable, and no issues or problems have been encountered as of now, as it is in production.
Dify's scalability is good and it handles growth or increased workloads effectively, depending upon the resources available. If server capacity is increased, Dify scales accordingly.
Dify's customer support has not been directly contacted as GitHub issues and the community have helped with any issues faced.
Negative
N8n was previously used before Dify.
Dify was chosen over n8n because Dify has more marketplace, plugins, and integrations than n8n, and it is lightweight compared to n8n, which is resource-heavy.
Dify was not purchased through the AWS Marketplace but rather the Git repository was cloned.
Dify is free to use and has a free license from GitHub under a Dify open-source license based on Apache 2.0, so no hurdles have been felt in deploying it or using it.
The automation work has saved time. For example, if a task would require an hour, it can now be done in seconds, so the time saved varies depending on the task. Exact numbers are not available, but it has saved a lot of time and also money, reducing the manual work that was done earlier.
Dify is free to use and has a free license from GitHub under a Dify open-source license based on Apache 2.0, so no hurdles have been felt in deploying it or using it.
N8n, Make.com, and Zapier were evaluated before choosing Dify. Make.com and Zapier are paid and Dify is absolutely free. Since n8n is also free but a little more complex than Dify, Dify was chosen over any other platform.
Dify should be considered by others looking into using it if they want a lightweight platform with feature-rich plugins to integrate their workflow and manage multiple automation tasks using a single platform. This is the best option available. The overall review rating for Dify is 9 out of 10.
We used Dify to create and test an agentic workflow and an AI agent model with some of the tools and RAG models. We used it to test how it works and how to implement it for part of our core product in our company.
We created an AI marketing agent using Dify, and the idea was that it can look into your marketing platforms, for example, Facebook, Google Ads, and Google Analytics. Those are the marketing platforms that we targeted, and instead of manually moving through dashboards, the idea was an agent will have access to your marketing data and can go through this data and provide you with reports and insights and suggestions. We used Dify's visual workflow builder to build this.
I tested Dify about six months ago for some of the tasks that we had for building some kind of a product, and I used it for two months and then did not use it very much after that.
The visual workflow builder Dify offers is really helpful, and you do not have to code everything. You can use it to connect nodes and make a flow of how your agent should work. Dify has RAG functionality that is also in-built, and those are the features that we used, and those two features were very good.
The visual workflow builder made my work easier because it saved us time since we did not have to code everything. One of the other interesting things was that it was really easy to show or present to a non-technical person. Our CEO was non-technical, so for him, it was really easy to show it as a diagram and explain how it works, and he could even do some edits. The ability for non-technical people to look into it is a really great use case.
Using Dify has positively impacted our organization because we were able to cut down on some development time and do a lot of testing in a very small time period. Initially, we had about two weeks of time to implement the whole thing, but that was cut down to two days of time through using Dify.
My personal experience with Dify's UI is that it is not my favorite, as it can be improved a little bit, and sometimes the UI feels a little bit buggy. I am not sure if that is because it was a self-hosted version. The documentation can also be improved a little bit more. I think not a lot of people are using Dify currently, so that is why the documentation is not very great. If the documentation was improved, that would also be a really good thing.
Currently, Dify could improve by offering better observability like other platforms. We currently use OpenAI Agents SDK, which requires you to build everything by code, but the observability is really good. It has OpenAI Traces, and you can basically trace everything for a conversation. If Dify had that kind of tracing functionality, that would be great.
I actually tested Dify about six months ago for some of the tasks that we had, it was for building some kind of a product, And, I used it for like two months and then, did not use it very much after that.
Dify is stable.
We used Dify only for the POC, so we did not expose it to a lot of workload. One of the main concerns that we had is that it might not be very scalable because we are hosting it in a self-hosted environment, and we have to configure the architecture and everything. Rather than using a cloud-hosted platform, using a self-hosted platform means there can be scalability issues. We anticipated there would be scalability issues, but we did not go for that scale. While testing, sometimes because of the limitations of the server, it crashed, stopped working, or got delayed, so it has a little bit of scalability problems.
We used Dify for testing out a POC and different ways of how to implement the agent, so there is no direct return on investment, as the investment was zero, so there is no return.
My experience with pricing, setup cost, and licensing is that it was free to use. We were able to get the free license from the GitHub release and then deploy it in our organization.
We evaluated another option called Chatbot Kit, but I am not very sure about that because I do not remember everything. We also used another product, either Chatbot Kit or ManyChat, and then after Dify, we switched back to OpenAI Agents SDK.
The visual workflow builder made my work easier because it saved us time since we did not have to code everything. One of the other interesting things was that it was really easy to show or present to a non-technical person. Our CEO was non-technical, so for him, it was really easy to show it as a diagram and explain how it works, and he could even do some edits. The ability for non-technical people to look into it is a really great use case.
Dify is self-hostable, so we did not have to pay anything. We just had to host it, and we really own the whole thing, and we can see the code as well. Self-hostability is another great feature.
If you are a startup or someone who is trying to run a POC related to agents, you should use Dify. It is a really good alternative that you can use to test things out and build a POC. After that, make sure to move to a different platform because if you need to scale it up and if you need custom steps, that is the advice I would have. I would rate this review a six out of ten.