Test Sifu Chat

Sifu Chat allows one to add a chatbot to any website. Below is an option. I still need to play around with it a little more but. But it lets one add their own data to train their bot. Part 2 will include more information on my tests.

Get your own!!!

Workflow Automation Agents: A Guide to AI and Vibe Coding Part 1

We have several tools to avail ourselves of to generate software solutions to take care of rudimentary busy work in our work flows. Through Generative AI and LLM technology, we have unprecedented opportunities to do more faster and focus on the big picture of projects without wasting time on minute details.

The goal of this endeavor is to compare and document some of the popular ones: n8n, Make.com, Zapier, Pokee AI, and Pipedream.

What is "Vibe Coding"?

Before we dive in, it’s important to define our approach. We are engaging in Vibe Coding: using AI to handle the syntax, boilerplate, and technical "how-to," allowing us to focus purely on the "vibe"—the flow, logic, and desired outcome of the application. The AI writes the code; we direct the symphony.

The Philosophy of the Tools

I have learned a lot about trying to solve the same problem with different tools. Each one has strengths and weaknesses. Based on my testing, here is how the landscape divides:

• Zapier: The "Apple" of automation. It is the most powerful in terms of being able to generate everything you need from a prompt. It is expensive but offers the easiest "plug-and-play" experience.
• Make.com (formerly Integromat): The "Visual Engineer." It offers a digital whiteboard where you can see logic branches and loops clearly. It is often cheaper than Zapier but has a slightly steeper learning curve.
• n8n: The "Power User’s Choice." It allows for self-hosting and infinite flexibility. It uses a node-based system that feels very close to writing code but with visual guardrails.
• Pipedream: The "Developer’s Speed Tool." It focuses on serverless event handling. If you know a little Python or JavaScript, this is often the fastest way to build custom endpoints.
• Pokee AI: An emerging AI-first platform that attempts to flatten the curve even further by relying heavily on natural language processing to build flows.

---

Case Study 1: YouTube to Blog Automation

The Goal: Create an automation that loops through at most 30 videos in a playlist on YouTube. For each video, the workflow must:

1. Transcribe the video.
2. Use AI to summarize the video.
3. Create a draft on Blogger containing the title, thumbnail, summary, and video embed.
4. Log the video and blog post in a MySQL database.
5. Send a notification via WhatsApp or Slack.

The Logic:

I used the same logic for all tools: Get the videos from the playlist, loop through them, transcribe via code or module, summarize via LLM, compose the draft, and notify the user.

The Implementation Results

Below is the comparison of how each platform handled this specific workflow.

Table 1.1: Automations for Converting YouTube Videos into Blog Posts

Platform	Obtain Videos	Transcribe	Summarize	Construct Blog Post	Log to MySQL	Notify	Difficulty (1-10)	Prompting Strategy
Zapier	Native Trigger (New Video in Playlist)	OpenAI Whisper (via Zapier integration)	ChatGPT (Conversation Action)	Blogger Action (Create Post)	MySQL Action (Insert Row)	Slack Action (Send Channel Message)	2/10	High. Zapier guided me through the whole process. I could edit the model through prompts, adding or removing steps easily.
n8n	YouTube Node (Get Playlist Items)	Deepgram/OpenAI Node	LangChain / AI Agent Node	HTTP Request (Blogger API)	MySQL Node	Slack Node	6/10	Medium. Required setting up specific nodes. The "Loop" node is powerful but requires understanding data structure (JSON).
Make	YouTube Watcher	Whisper Module	OpenAI Module	Blogger Module	MySQL Module	Slack Module	4/10	Medium. The visual "Iterator" makes looping through 30 videos very intuitive, but error handling requires extra modules.
Pipedream	YouTube Data API (Source)	Python Code Step	OpenAI API	Blogger API (Auth required)	MySQL Resource	Slack API	7/10	Low. Relied heavily on code generation. Excellent for "Vibe Coding" if you can read the Python it generates.
Pokee AI	AI Agent	AI Agent	AI Agent	AI Agent	AI Agent	AI Agent	3/10	High. Relies almost entirely on the prompt to configure the agents. Very "black box" but effective for standard flows.
Python	google-api-python-client	speech_recognition	openai lib	blogger_v3 API	mysql-connector	slack_sdk	9/10	None. Pure coding. Requires handling OAuth2.0 tokens manually for YouTube and Blogger, which is the hardest part.

   Screenshot of the Zapier Flow showing the linear progression

Screenshot of the n8n Workflow showing the complex loop node

The "Mega Prompt" Used

For the AI-driven tools (Zapier and Pokee AI), this is the prompt that successfully generated the majority of the workflow:

"Create an automation that triggers when a new video is added to my specific YouTube playlist. Loop through the videos (limit to 30). For each video, use OpenAI to transcribe the audio and then summarize the content into a blog post format. Create a new draft post on my Blogger account with the Video Title as the headline, the thumbnail as the header image, and the summary as the body. Embed the video at the end. Finally, add a row to my MySQL database with the video ID and blog URL, and send a Slack message to the '#content' channel confirming the draft is ready."

---

Case Study 2: X.com Bookmarks to Video Archive

The Goal: Loop through bookmarked X.com posts, download the video (if present), log it in MySQL, and notify via Slack/WhatsApp.

Table 1.2: Automation Difficulty & Implementation

Platform	Difficulty	Key Challenge
Zapier	5/10	Native "Download Video" actions are rare; often requires a third-party helper tool like "CloudConvert" or a custom webhook.
n8n	4/10	Excellent HTTP Request handling makes downloading binary files (videos) and uploading them to storage easier.
Make	4/10	Similar to n8n, handles binary file manipulation well.
Python	6/10	yt-dlp or similar libraries make this easy in Python, but hosting the script and storage requires a server.

---

Case Study 3: Social Media Repurposing

The Goal: Select a previous blog post, summarize it, and cross-post to Facebook, X, Instagram, and TikTok.

Table 1.3: Automation Difficulty & Implementation

Platform	Difficulty	Key Challenge
Zapier	3/10	Very strong social media integrations. Posting to TikTok/Instagram usually requires a business account connection.
Make	3/10	Great for "router" logic (e.g., "If post has image -> Instagram; If text only -> X").
Pipedream	5/10	managing 4 different OAuth tokens for social platforms can be tedious in code.

---

Tools and Platforms Overview

To wrap up, here are the specific tools I used for these experiments.

Table 2.0: Platform Feature Matrix

Tool	Cost	No Code	Low Code	AI Support	Custom Code	Database Int.
n8n	Free (Self-hosted) / $20 Cloud	✅	✅	✅	✅ (JS/Python)	✅
Make	Free / $9/mo	✅	✅	-	-	✅
Zapier	Free / $19.99/mo	✅	-	✅	✅ (Python steps)	✅
Pipedream	Free / $19/mo	-	✅	✅	✅ (Node/Python)	✅
Pokee AI	Varies	✅	-	✅	-	✅

Software & APIs Used:

• Google Gemini / OpenAI: For summarization and intelligence.
• YouTube API: For retrieving video content.
• Blogger API: For content publishing.
• Aiven: Managed MySQL database hosting.

Check Out this Video

Conclusion

If you want speed and don't mind the cost, Zapier is the undisputed king of "Vibe Coding"—it just works. If you want to build a complex "software" backend without maintaining a server, n8n and Make offer the visual control you need. For those who want to own every line of logic, Python (aided by Pipedream) remains the gold standard.

(Future updates will include detailed code snippets for the Pipedream and Python implementations.)