AccessLevel
Free

Source
https://github.com/svcvit/Awesome-Dify-Workflow/blob/main/DSL/translation_workflow.yml

Testing Environment
Dify Cloud

LLMs Used
gpt-4o

ModelType
Text-Only;

IsFunctional
Yes

Performance Rating
Great

 Expected Behaviour
Enable High-Quality Translations with Language and Country-Specific Adaptations Based on Input Text

Actual Behaviour
After thorough testing, the translation quality is exceptionally high and closely aligned with the original text. The results are particularly impressive when a specific country is selected, as it further refines the translation to match the nuances and style of the local language

Evaluation
AI Native: (9/10) The workflow demonstrates excellent integration of AI capabilities, especially in leveraging context-sensitive and adaptive translation, perfectly aligning with AI Native principles.

Workflow Breakdown
1⃣ Input the Text for Translation
• Enter the text you wish to translate. Select the desired language and country to ensure the translation aligns with the specific linguistic and cultural context. The system will automatically identify the key content.

2⃣ Identify Technical Terms
• Scan the text to detect any specialized or technical terms that may require precise handling.

3⃣ Perform Direct Translation
• Translate the text directly using the initial translation engine, providing a basic understanding of the content.

4⃣ Analyze Translation Issues
• Review the direct translation for any inaccuracies or misinterpretations, particularly with technical or contextual terms.

5⃣ Second Translation Based on Meaning
• Reinterpret problematic areas by analyzing their intended meaning and refining the translation accordingly.

6⃣ Generate Final Translated Text
• Combine the refined translations and produce the final output, ensuring accuracy and fidelity to the original meaning.

Statement: Evaluation results are generated by AI, lack of data support, reference learning only.

That’s all for the case study. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #20 – dify – Building a High-Quality Translation Workflow with Dify: Inspired by Andrew Ng’s Walk and Translate Approach appeared first on AI Native Foundation.

AccessLevel
Free

Source
https://github.com/svcvit/Awesome-Dify-Workflow/blob/main/DSL/SEO Slug Generator.yml

Testing Environment
Dify Cloud

LLMs Used
gpt-4o

ModelType
Text-Only;

IsFunctional
Yes

Performance Rating
Decent

 Expected Behaviour
Generate a Clean and SEO-Friendly URL Slug for Your Blog Post Based on the Input Title

Actual Behaviour
After testing with title “Building a High-Quality Translation Workflow with Dify: Inspired by Andrew Ng’s Walk and Translate Approach”, the slug high-quality-translation-workflow-dify-andrew-ng is concise, SEO-friendly, and effectively captures key aspects of the title. However, removing “Andrew Ng” could make it more focused, such as high-quality-translation-workflow-dify.

Evaluation
AI Native: (7/10) The workflow leverages AI to automate slug generation but could further improve efficiency and personalization to fully align with AI Native principles.

Workflow Breakdown
1⃣ Input Text for Slug Generation
• Provide the text or title for which you want to create a slug.
2⃣ Optimize and Generate Slug with GPT
• Use GPT to refine the input and generate a concise, SEO-friendly slug tailored to the text.

Statement: Evaluation results are generated by AI, lack of data support, reference learning only.

That’s all for the case study. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #19 – dify – SEO Slug Generator appeared first on AI Native Foundation.

AccessLevel
Free

Source
https://cloud.dify.ai/explore/apps?category=Workflow

Testing Environment
Dify Cloud

LLMs Used
gpt-3.5-turbo

ModelType
Text-Only;

IsFunctional
Yes

Performance Rating
Decent

 Expected Behaviour
Workflow for retrieving information from the internet, followed by segmented generation of SEO blogs.

Actual Behaviour
The generated results are quite satisfactory. The workflow effectively summarizes the outcomes of Google keyword searches and creates blog-style content based on the findings.
Strengths:
• Text-Based Summaries: The workflow provides clear and coherent summaries, making it suitable for generating written content quickly.
Limitations:
1. No Source Links: The summarized blog lacks links to the original articles or references, which could impact credibility and traceability.
2. No Images: The output is entirely text-based, without any accompanying visuals, which might reduce engagement and readability.
Conclusion:
While the summarization quality is good, adding features like source attribution and relevant images would greatly enhance the utility and appeal of the generated content.

Evaluation
AI Native: (7/10) The workflow demonstrates strong AI integration but lacks key features like source attribution and multimedia support, which are crucial for a comprehensive AI Native experience.

Workflow Breakdown
1⃣ Collect User Input
• Function: Gather essential details from the user, including:
• Keyword: The main keyword for the article.
• Title: The article title (optional).
• Audience: The target audience for the content.
• Words/Brands to Avoid: Specific terms or brands to exclude.
• Tone: The desired tone of the article (e.g., professional, casual).
2⃣ Check Title Validity
• Condition: Determine if the title input is empty.
• If Empty: Automatically generate an SEO-optimized title.
• If Provided: Use the given title and proceed.
3⃣ Generate Title and Related Keywords
• Function: Based on the user-provided keyword:
• Generate an SEO-optimized title (if necessary).
• Identify and generate related keywords for enhancing search relevance.
4⃣ Google Search for Information
• Setup: Ensure a third-party API key is configured with Serp.
• Function: Use the Google Search tool to gather accurate and comprehensive information for content generation.
5⃣ Generate Outline and Article
• Process:
• Create a detailed article outline based on the search results and user inputs.
• Generate the introduction and main body, ensuring the content aligns with the specified tone, audience, and exclusions.
6⃣ Transform and Output
• Function:
• Merge the introduction and main body to form a complete article.
• Output the finalized content in a user-ready format.
This structured process ensures the creation of accurate, SEO-friendly, and tailored content that meets user requirements.

Statement: Evaluation results are generated by AI, lack of data support, reference learning only.

That’s all for the case study. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #18 – dify – SEO Blog Generator appeared first on AI Native Foundation.

AccessLevel
Free

Source
https://cloud.dify.ai/explore/apps?category=Workflow

Testing Environment
Dify Cloud

LLMs Used
gpt-3.5-turbo

ModelType
Text-Only;

IsFunctional
Yes

Performance Rating
Decent

 Expected Behaviour
Based on users’ preferences, retrieve external knowledge to create more precise article summaries.

Actual Behaviour
After testing, this workflow provides two forms of summarization:
1. Knowledge Base Retrieval: Used for generating technical summaries by retrieving information from a knowledge base.
2. GPT-Based Summarization: Uses GPT to generate summaries directly.
Observations:
• Input Length Limitation: The default input length is restricted to 200 characters, which requires manual modification to handle longer inputs.
• Output Length Customization: The workflow does not allow setting a specific output length, which also requires manual adjustments to the configuration.
Despite these limitations, the summarization results are quite accurate and effective for most use cases. However, improving flexibility in input and output length configurations would make the workflow more user-friendly.

Evaluation
AI Native: (7/10) Decent integration with AI Native, but limited input/output flexibility restricts user adaptability and scalability.

Workflow Breakdown
1⃣ Input the Text for Summarization
• Enter the text you want to summarize. The system will process the content for further steps.
2⃣ Choose Summarization Type
• Technical Summary: Focus on generating precise technical insights.
• General Overview: Provide a high-level summary of the content.
3⃣ Process Based on Selection
• For Technical Summary:
• Search the knowledge base for relevant information.
• Use the knowledge base data to generate an accurate and detailed technical summary.
• For General Overview:
• Utilize GPT to analyze the input text.
• Generate a concise and coherent summary based on overall content.
4⃣ Format and Output the Summary
• Convert the summary into the desired format (e.g., bullet points, paragraphs).
• Present the summarized content for review or further use.

Statement: Evaluation results are generated by AI, lack of data support, reference learning only.

That’s all for the case study. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #17 – dify – Text Summarization Workflow appeared first on AI Native Foundation.

AccessLevel
Free

Source
https://cloud.dify.ai/explore/apps?category=Workflow

Testing Environment
Dify Cloud

LLMs Used
chatgpt-4o

ModelType
Text-Only;

IsFunctional
Yes

Performance Rating
Decent

 Expected Behaviour
Workflow for enhancing translation accuracy by identifying proper nouns, literal translations, pointing out issues with literal translations, and suggesting idiomatic translations.

Actual Behaviour
After testing, this workflow successfully translates English into Chinese using a three-step translation method. The translation quality is quite close to the original text and is fairly good. However, there are some limitations:
1. Language Selection: It appears that the workflow only supports English-to-Chinese translation and does not allow selecting other languages.
2. Country Selection: There is no option to specify a country for localized translation preferences.
3. Customization: Manual modifications to the workflow are required to enable translations for other language pairs.
Overall, while the translation quality is satisfactory, the workflow could benefit from added flexibility in language and localization options.

Evaluation
AI Native: (7/10) The workflow demonstrates strong automation and idiomatic understanding, but lacks multilingual and localization flexibility, which limits its AI Native compatibility.

Workflow Breakdown
1⃣ Input the Text for Translation
• Enter the text you wish to translate. The system will automatically identify the key content.
2⃣ Identify Technical Terms
• Scan the text to detect any specialized or technical terms that may require precise handling.
3⃣ Perform Direct Translation
• Translate the text directly using the initial translation engine, providing a basic understanding of the content.
4⃣ Analyze Translation Issues
• Review the direct translation for any inaccuracies or misinterpretations, particularly with technical or contextual terms.
5⃣ Second Translation Based on Meaning
• Reinterpret problematic areas by analyzing their intended meaning and refining the translation accordingly.
6⃣ Generate Final Translated Text
• Combine the refined translations and produce the final output, ensuring accuracy and fidelity to the original meaning.

 Statement: Evaluation results are generated by AI, lack of data support, reference learning only.

That’s all for the case study. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #16 – dify – Three-Step Translation Workflow appeared first on AI Native Foundation.

AccessLevel
Free

Source
https://cloud.dify.ai/explore/apps?category=Workflow

Testing Environment
Dify Cloud

LLMs Used
chatgpt-4o-latest

ModelType
Text-Only;

IsFunctional
Yes

Performance Rating
Decent

 Expected Behaviour
Based on the input address, it can help me summarize the main points of the article and the address of related pictures or videos

Actual Behaviour
Through testing, the workflow efficiently summarizes the input address into key themes and points. However, it currently lacks the capability to extract relevant image URLs. Additionally, Jina Reader is a paid tool with a limited number of free trial uses.

Evaluation
AI Native: (7/10) The workflow efficiently utilizes AI for summarization, meeting AI Native goals, but lacks image URL extraction and relies on a paid tool, hindering accessibility and scalability.

Workflow Breakdown
1⃣ Provide the website URL
• Enter the URL of the website or article you’d like to summarize. The system will automatically crawl the content for you.
2⃣ Summarize and Analyze
• Using GPT, the system will summarize the crawled content and extract two key elements:
• Discussion Topics: Highlight the main themes or points of debate.
• Key Takeaways: Identify the most important insights or actionable information.

Statement: Evaluation results are generated by AI, lack of data support, reference learning only.

That’s all for the case study. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #15 – dify – Summarize website with Jina Reader appeared first on AI Native Foundation.

Source
https://cloud.dify.ai/explore/apps?category=Workflow

Testing Environment
Dify Cloud

LLMs Used
gpt-3.5-turbo

ModelType
Text-Only;

IsFunctional
Yes

Performance Rating
Great

 Expected Behaviour
A multifunctional email assistant capable of summarizing, replying, composing, proofreading, and checking grammar.

Actual Behaviour
After evaluation, the multifunctional email assistant effectively handles summarizing, replying, composing, proofreading, and grammar checking. It performs well across these functions; however, the current implementation uses GPT-3.5, which is relatively outdated. Upgrading to GPT-4.0 is recommended to achieve improved performance and more advanced capabilities.

Evaluation
AI Native: (9/10) The workflow demonstrates high alignment with AI Native principles by effectively leveraging advanced AI (GPT) for automation and customization, although an updated model upgrade to GPT-4 would enhance its adaptive capabilities.

Workflow Breakdown
1⃣ Input Your Email
• Begin by entering the email content. Then, choose one of the available options: Summarize, Reply, Compose, or Improve. This selection determines the next action.
2⃣ Process with GPT
• Based on your chosen option, the system uses GPT to perform the selected task:
• Summarize: Extract key points and provide a concise summary of the email.
•Reply: Generate a thoughtful and relevant reply.
•Compose: Draft a new email based on your input and requirements.
•Improve: Refine the email by enhancing clarity, grammar, and tone.
3⃣ Transform the Output
• The result generated by GPT is processed through a template conversion tool to match the desired format or style, ensuring the output meets your specific needs.

Statement: Evaluation results are generated by AI, lack of data support, reference learning only.

That’s all for the case study. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #14 – dify – Email Assistant Workflow appeared first on AI Native Foundation.

Source
https://cloud.dify.ai/explore/apps

Testing Environment
Dify Cloud

LLMs Used
chatgpt-4o-latest

ModelType
Text-Only;

IsFunctional
Yes

Performance Rating
Needs Improvement

 Expected Behaviour
Through this workflow, the user can start by entering a question, use Tavily Search to search, Jina Reader to crawl web content, and generate summaries, ultimately obtaining a structured table of results.

Actual Behaviour
After thorough testing, the workflow demonstrates the ability to generate a structured result table based on input keywords. It successfully extracts the website URL and a summary of the content; however, it falls short in the following areas:
1. Missing Image Addresses: While the workflow provides website URLs and summaries, it does not retrieve corresponding image addresses, which could enhance the usability of the output for visual content needs.
2. Limitations in Recent News Retrieval: The workflow is unable to search for and summarize news from the latest 48 hours, which limits its relevance for time-sensitive use cases.
3. Duplication Bug: There is a notable bug where each piece of generated data appears duplicated in the final output, leading to inefficiencies and potential confusion.

Recommendations
1. Enhance Image Address Retrieval: Implement functionality to extract image URLs alongside summaries to improve the completeness of the results.

2. Expand News Search Capabilities: Include a mechanism to access and process news articles published within the last 48 hours for real-time relevance.

3. Fix Duplication Bug: Address the issue causing duplicated entries to ensure cleaner and more reliable outputs.

Evaluation
AI Native: (7/10) While the workflow demonstrates structured automation and integrates AI tools, issues like duplication, missing image data, and news retrieval limitations hinder full AI Native potential.

Workflow Breakdown
1⃣ Ask a Question
– Start by entering a question related to the topic you want to explore. This question will guide the search and retrieval process.

2⃣ Search with Tavily
– The system uses Tavily Search to find relevant web pages and articles that align with your question. It ensures the results are comprehensive and contextually relevant.

3⃣ Crawl Web Content with Jina Reader
– The identified web pages are processed through Jina Reader, which crawls the content and extracts key information for analysis.

4⃣ Summarize the Results
– Using GPT, the system generates concise summaries of the crawled content, focusing on the most critical points and themes.

5⃣ Generate a Structured Table
– The summarized information is organized into a structured table, making it easy to review key findings, including website URLs, summaries, and other relevant data.

Statement: Evaluation results are generated by AI, lack of data support, reference learning only.

That’s all for the case study. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #13 – dify – Web Content Search and Summarization Workflow appeared first on AI Native Foundation.

Workflow Breakdown
1⃣ Input Key Info: Provide Keyword, Title, Audience, Brand to Avoid, Tone, and Voice.
2⃣ GoogleSearch Integration: Fetch relevant content through Keyword and Title searches.
3⃣ GPT Generation: Summarize and analyze the content tailored to Audience, Brand to Avoid, and Tone & Voice.
4⃣ Template Conversion: Convert GPT-generated content into blog format using a template tool.

Insights
Dify’s SEO Blog Generator makes it easy to create high-quality blogs efficiently! However, the lack of related article links and images is a downside. Adding these would make it a flawless tool!

That’s all for the workflow insights. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #5 – Dify – Unlocking SEO Automation: My Experience with Dify SEO Blog Generator Workflow appeared first on AI Native Foundation.

That’s all for the workflow insights. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.

The post AI Native Flow Case Study #2 – Dify – Trending News Discovery and Summarization appeared first on AI Native Foundation.