2. Make funny birthday cards for niche

I want to build out a task that makes funny birthday cards for any niche. I would input a description of my niche audience or topic, and the final output would be a collection of completed images (in jpeg), with funny text and image. I think we could generate the image with any of our AI generators, eg Dalle, and provided we prompt something that looks hard drawn the end result would be good. But what we really want to spend time on is the writing of the puns for the card. These should be funny and relevant to the niche/topic in question. We would therefore want a series of prompts which researches the topic in question, looks for possible words and phrases that are relevant to the niche, and then generates a list of slogans for that niche. Since this is such an important step we likely want to run a few searches to collect possible names, and then a few LLM prompts that import a summary of these search results and use the research to generate slogan ideas. We are likely to need some custom prompts to generate "play-on" words. The slogans are important. We should likely run these prompts a few times, and then have another AI review and choose the best slogans. And then it's just a case of inserting the text and images into a layout. I have a tool called Bannerbear which lets me build template image structures, so I will likely make 2-3x of these before we formalize the task, and then we will just slot "card-text-line-1" and "card-text-line-2" and "doodle-image-url" into this template to give us our cards. These can be 1000x1000 to start (Bannerbear lets me scale up easily). Research is very important with this one: before we start the whole process, we should probably search Google images (for "{niche topic} birthday cards") and see what current top-selling images in the niche look like, and then from there we can begin to grab some relevant niches.

Sub Agent #1 | Sub Agent #2 | Sub Agent #3

final-output-comps..1

Great question! Let's break down the final output and the components required to achieve the automated workflow for creating funny birthday cards for any niche.

a) Final output:
The final output of this automated workflow would be a collection of completed birthday card images in JPEG format. Each card would feature a funny, relevant slogan and a doodle-style image related to the niche or topic provided as input. The dimensions of the images would be 1000x1000 pixels, but they should be easily scalable using the Bannerbear tool.

b) Core components/outputs:
1. Niche/topic input: A text input that describes the target audience or topic for the birthday cards (e.g., "fitness enthusiasts" or "cat lovers").

2. Research data: JSON data containing relevant words, phrases, and information about the niche/topic. This data would be generated by running multiple searches and collecting the results.

3. Slogan options: A list of potential slogans (in text format) generated by the LLM using the research data. These slogans should be funny, relevant to the niche, and may include play-on-words or puns.

4. Best slogans: A refined list of the best slogans (in text format) selected by another AI from the generated slogan options.

5. Doodle-style images: URLs of doodle-style images related to the niche/topic, generated using an AI image generator like DALL-E. These images should have a hand-drawn appearance to match the desired card style.

6. Bannerbear templates: 2-3 pre-designed image templates in Bannerbear, which will be used to structure the birthday cards. These templates should have placeholders for "card-text-line-1", "card-text-line-2", and "doodle-image-url".

7. Final birthday card images: The completed birthday card images in JPEG format, generated by inserting the best slogans and doodle-style images into the Bannerbear templates. Each card should be 1000x1000 pixels in size.

To summarize, the core components required for this automated workflow include the niche/topic input, research data, slogan options, best slogans, doodle-style images, Bannerbear templates, and the final birthday card images. The process involves researching the niche, generating and selecting slogans, creating relevant images, and combining these elements using the Bannerbear templates to produce the final output.

final-output-comps..2

## Understanding the Final Output of the AI-Generated Birthday Card Workflow

### a) Final Output

The final output of this workflow would be a set of **digital birthday cards in JPEG format**. Each card would feature:

*   **A humorous slogan or pun**: This text would be directly related to the chosen niche and generated through AI with a focus on wordplay and relevant humor.
*   **An AI-generated image**: This image, likely created with tools like Dall-E, would visually represent the niche in a "hand-drawn" style, enhancing the comedic effect.
*   **A pre-designed layout**: The text and image would be placed within a consistent template (created in Bannerbear) to ensure a polished and visually appealing final product.

The output essentially provides several ready-to-use birthday cards for a specific niche, catering to a particular audience with inside jokes and relevant humor.

### b) Core Components

To achieve the final output, several key components are needed:

*   **Niche Research**:
    *   **Keywords and Phrases**: A list of relevant terms and phrases associated with the chosen niche, gathered through search engine queries and analysis of existing birthday cards within that niche. 
    *   **Image Inspiration**: A collection of top-selling birthday card images for the niche, sourced from platforms like Google Images, to understand current trends and visual styles.
*   **Slogan Generation**:
    *   **AI Prompts**: A set of prompts designed to generate puns and humorous slogans related to the niche, incorporating the researched keywords and phrases.  This might involve several iterations and refinements to achieve optimal results.
    *   **Slogan Selection**: An AI model or algorithm to evaluate and select the best slogans based on humor, relevance, and originality.
*   **Image Generation**:
    *   **AI Image Prompts**: Precise prompts for tools like Dall-E to generate images that visually represent the niche and complement the chosen slogan, maintaining a "hand-drawn" aesthetic.
*   **Card Layout**:
    *   **Template Design**: Creation of 2-3 templates in Bannerbear, defining the layout structure for text placement, image placement, and overall design elements. 
    *   **Text and Image Integration**:  Combining the selected slogan and generated image within the chosen template, ensuring proper formatting and visual balance.

By breaking down the process into these components and utilizing AI effectively at each stage, you can build a robust workflow for generating humorous and niche-specific birthday cards.

final-output-comps GPT..3

a) The final output of your task would be a collection of images, specifically humorous birthday cards tailored to unique target niches. These images, which will be in JPEG format, will be a combination of funny text and AI-generated artwork. Each of these cards will be based on a niche specified as an input and will feature puns and jokes that relate to that niche.

b) The process of creating this final output can be divided into various components:

1. Niche Research: One of the first steps would be to gather information about the niche for which the card is to be created. This might involve generating a list of relevant phrases, words and context for the specific niche. This can be done by utilizing AI algorithms to search and scrape this data from the Internet.

2. Creation of Slogans and Puns: Based on the research conducted, the next step would involve generating funny texts, slogans and puns that resonate with the niche. The AI agent would have to create multiple variants of these texts, perhaps through the use of language model algorithms, and seek to make them humorous and appropriate for the given niche.

3. Image Generation: Concurrently, the AI agent would use one of the AI generators like Dalle to create relevant and funny images tailored to the given niche. This could be optimized by prompting the AI to produce images that look hand-drawn.

4. Evaluation of Texts: The AI agent would then review all the generated slogans and puns, select the best ones. This could also involve refining these texts based on the AI judgment or further iterations of generation and review.

5. Integration into Card Layout: Once the texts and images are ready and reviewed, they are then inserted into the card template created on Bannerbear. This would mean placing the text within "card-text-line-1" and "card-text-line-2" and assigning the generated image on the "doodle-image-url" spot.

6. Scaling of Final Output: Finally, the completed cards would be scaled to a size of 1000x1000, using the functionality offered by Bannerbear.

7. Competition Analysis: Before starting the process, a brief analysis of what top-selling niche birthday cards look like on the market is done for inspiration and better targeting. AI can scrap this data from the Internet, typically from Google images.

Each of these components will require different data types and processing. For instance, text data will be needed for niche research and pun creation, while image data will be needed for the use with AI image generators. Scripts or specific commands may be required to automate the integration of these elements into the Card Layout on Bannerbear.

CLAUDE:- final-output-and-component-outputs

a) The final output of this automated workflow would be a collection of funny birthday card images in JPEG format, each tailored to a specific niche or topic. These images would feature:

1. Humorous slogans or puns relevant to the niche, generated through AI-powered research and wordplay.
2. AI-generated doodle-style images that visually represent the niche and complement the chosen slogans.
3. A pre-designed layout created using Bannerbear templates, ensuring a polished and visually appealing final product.

The dimensions of the images would be 1000x1000 pixels, but they should be easily scalable using the Bannerbear tool.

b) To achieve the final output, the following core components and outputs are required:

1. Niche/Topic Research:
   - Output: A list of relevant niches or topics for which birthday cards will be created.
   - Process: Perform Google image searches for "{niche topic} birthday cards" to identify popular and relevant niches. Gather information about each niche, including relevant keywords, phrases, and visual inspiration from top-selling cards.

2. Slogan Generation:
   - Output: A list of humorous slogans or puns for each niche, generated based on the niche-specific research.
   - Process: Use custom LLM prompts that take the niche-specific keywords and phrases as input and generate multiple slogan ideas, including play-on-words and puns. Run these prompts multiple times to generate a diverse set of options.

3. Slogan Selection:
   - Output: A curated list of the best slogans for each niche.
   - Process: Employ another AI model to review and select the most humorous and relevant slogans from the generated options.

4. Doodle Image Generation:
   - Output: AI-generated doodle-style images relevant to each niche.
   - Process: Use AI image generation tools like DALL-E to create doodle-style images based on the niche and selected slogans. Ensure the images maintain a hand-drawn aesthetic.

5. Birthday Card Template Creation:
   - Output: 2-3 birthday card layout templates in Bannerbear, with placeholders for text and images.
   - Process: Design and set up birthday card templates in Bannerbear, with specific areas for "card-text-line-1", "card-text-line-2", and "doodle-image-url".

6. Final Birthday Card Image Generation:
   - Output: A collection of funny birthday card images in JPEG format, sized at 1000x1000 pixels.
   - Process: Combine the selected slogans and generated doodle images with the Bannerbear templates to create the final birthday card images for each niche. The slogans will be inserted into the "card-text-line-1" and "card-text-line-2" placeholders, while the doodle images will be added to the "doodle-image-url" placeholder.

Each component will require different data types and processing, such as text data for niche research and slogan generation, and image data for AI-generated doodles. Scripts or specific commands may be needed to automate the integration of these elements into the Bannerbear templates.

By executing these components in the specified order and leveraging AI effectively at each stage, you can build a robust workflow for generating humorous, niche-specific birthday cards that cater to particular audiences with relevant inside jokes and visually appealing designs.

claude 3 - extract final output

extract component outputs

To achieve the final output, the following core components and outputs are required:

Claude - research-ideas-and-input-variables

Thank you for the detailed description of the task. Let me address your questions one by one:

a) To clarify the final output, I would ask the following questions:
1. How many birthday card designs should be generated for each niche?
2. What file format and resolution should the final card images be in?
3. Should the cards be designed for print or digital use, or both?
4. Are there any specific design elements or styles that should be incorporated?
5. Should the AI agent also generate a list of the best slogans for each niche, or just use them in the card designs?

b) Before launching the task completion process, the AI agent should accept the following inputs:
1. A description of the niche audience or topic for which the birthday cards will be created.
2. The number of card designs to be generated for each niche.
3. The desired file format and resolution for the final card images.
4. Any specific design requirements or preferences, such as color schemes, fonts, or styles.

c) I have a few suggestions for segmenting the task into specialized areas:

1. Research and Analysis:
   - Conduct a separate search for relevant words, phrases, and themes related to the niche.
   - Analyze the top-selling birthday cards in the niche to identify common design elements and styles.

2. Slogan Generation:
   - Run multiple prompts to generate a diverse list of slogans for the niche.
   - Have a separate AI agent review and select the best slogans based on humor, relevance, and originality.

3. Image Generation:
   - Use AI image generators like DALL-E to create multiple doodle-style images relevant to the niche.
   - Have a separate AI agent curate and select the most appropriate images for the card designs.

4. Card Design and Assembly:
   - Create a few template structures in Bannerbear for the birthday cards.
   - Integrate the selected slogans and images into the templates to generate the final card designs.

By segmenting the task into these specialized areas, the AI agent can focus on each aspect individually, ensuring high-quality results at each stage before moving on to the next. This approach also allows for easier iteration and refinement of specific components if needed.

GPT - research-ideas-and-input-variables

a) To further understand the final output for the birthday card creation task, here are some potential questions:

- How many birthday cards are we expected to generate per niche or topic?
- What are some specifications for the completed images? Is there a particular style or color palette that should be followed?
- What is the length limit for the slogans or puns? Do they each need to fit a certain character or word count?
- Will there be a language or cultural considerations in creating the humor in the card?
- Should the images and slogans follow a specific tone of humor (e.g., lighthearted puns, sarcastic humor, etc.)?

b) Before starting the task completion process, the AI Agent would need these inputs:

- The niche or topic of interest for which the birthday cards are to be made.
- Examples or guidelines on the type of humor that is preferred for the niche.
- Details of the layout templates available in Bannerbear, including where the puns, slogans, and images should be slotted.
- Any stylistic or cultural references to be considered as part of the niche or topic.
- Parameters for content validation, such as how to determine the best slogans.

template-ideas

Based on the provided information, here are some template suggestions that could help refine the definition of a successful output for your funny birthday card generation task:

1. Etsy or other online marketplaces: Visit popular online marketplaces like Etsy and search for "funny birthday cards" or "{niche} birthday cards." Review the top-selling cards and their product descriptions to understand what makes them successful. Pay attention to the design elements, color schemes, and the types of humor used in the slogans.

2. Pinterest: Search for "funny birthday cards" or "{niche} birthday cards" on Pinterest. Analyze the most popular and visually appealing pins to identify common design trends, layouts, and humor styles that resonate with users.

3. Greeting card company websites: Visit the websites of popular greeting card companies like Hallmark, American Greetings, or Papyrus. Browse their collections of funny birthday cards and study the designs, slogans, and artwork that they use. These companies have a proven track record of creating successful cards.

4. Competitor analysis: Identify companies or individuals that create funny, niche-specific birthday cards. Visit their websites, social media profiles, or online stores to review their products. Analyze their design aesthetics, the types of humor they employ, and how they tailor their cards to specific niches.

5. Customer reviews: Read customer reviews for popular funny birthday cards on various platforms like Amazon, Etsy, or greeting card company websites. Pay attention to what customers appreciate about the cards, such as the cleverness of the slogans, the relatability of the humor, or the quality of the designs. Use these insights to guide your own card creation process.

6. Subreddits and online communities: Visit subreddits or online communities focused on humor, card-making, or specific niches. Look for posts showcasing funny birthday cards and read the comments to gauge audience reactions and preferences. This can provide valuable insights into what types of humor and designs resonate with different communities.

By collecting and analyzing these templates, you can develop a clearer understanding of what constitutes a successful funny birthday card for various niches. Pay attention to common design elements, humor styles, and how the cards effectively target specific audiences. Use these insights to guide your AI-powered card generation process, ensuring that your final output aligns with proven success factors in the greeting card industry.

claude 3 - searches

Here are 5 specific Google search keyphrases to help refine the definition of a successful output for the funny birthday card generation task:

1. "funny programmer birthday card etsy" - This search will yield results from Etsy, a popular online marketplace, showcasing top-selling funny birthday cards specifically tailored to programmers. Reviewing these cards will provide insights into design elements, humor styles, and slogans that resonate with this niche audience.

2. "best funny birthday cards for doctors pinterest" - Searching for the best funny birthday cards for doctors on Pinterest will reveal popular and visually appealing designs that cater to the medical profession. Analyzing these pins will help identify common design trends, layouts, and types of humor that work well for this specific niche.

3. "hallmark funny birthday cards for dog lovers" - Searching for funny birthday cards for dog lovers on the Hallmark website will provide examples from a well-established greeting card company. Studying these cards will showcase proven design aesthetics, slogans, and artwork that appeal to dog enthusiasts.

4. "funny golf birthday cards papyrus" - This search will yield results from Papyrus, another popular greeting card company, focusing on funny birthday cards for golf enthusiasts. Reviewing these cards will offer insights into how to effectively combine humor and golf-related themes to create successful niche-specific cards.

5. "funny birthday cards for teachers amazon reviews" - Searching for funny birthday cards for teachers on Amazon and reading customer reviews will provide valuable feedback on what customers appreciate about these cards. Pay attention to comments about the cleverness of the slogans, the relatability of the humor, and the overall design quality to guide your own card creation process.

By searching for these specific keyphrases and reviewing the top results, you'll be able to gather a collection of successful funny birthday card templates tailored to various niches. Analyze the common elements, humor styles, and design choices that make these cards effective, and use these insights to refine your own AI-generated cards, ensuring they align with proven success factors in the greeting card industry.

Funny Programmer Birthday Card on Etsy
Best Funny Birthday Cards for Doctors on Pinterest
Hallmark Funny Birthday Cards for Dog Lovers
Funny Golf Birthday Cards by Papyrus
Funny Birthday Cards for Teachers on Amazon Reviews

subagent-summary

A) FULL TASK SUMMARY: Create an automated workflow for generating funny, niche-specific birthday card images using AI-powered research, slogan generation, and image creation.

B) FINAL TASK OUTPUT: A collection of JPEG images, each sized at 1000x1000 pixels, featuring humorous slogans and AI-generated doodle-style images tailored to specific niches or topics.

C) COMPONENTS OVERVIEW:
1. Niche/Topic Research
2. Slogan Generation
3. Slogan Selection
4. Doodle Image Generation
5. Birthday Card Template Creation
6. Final Birthday Card Image Generation

D) OVERVIEW OF SUBAGENTS:
1. Niche Researcher: Conducts research to identify relevant niches and gather niche-specific information.
2. Slogan Generator: Generates humorous slogans and puns based on the niche-specific research.
3. Slogan Selector: Selects the best slogans from the generated options.
4. Doodle Artist: Creates AI-generated doodle-style images relevant to each niche.
5. Template Designer: Designs and sets up birthday card templates in Bannerbear.
6. Card Assembler: Combines the selected slogans, doodle images, and templates to create the final birthday card images.

E) SUBAGENT 1: Niche Researcher
- Name: Niche Researcher
- Final Output: A list of relevant niches or topics for birthday cards, along with niche-specific keywords, phrases, and visual inspiration.
- Tasks:
  1. Perform Google image searches for "{niche topic} birthday cards" to identify popular niches.
  2. Analyze search results to extract relevant keywords and phrases for each niche.
  3. Gather visual inspiration from top-selling cards in each niche.
  4. Compile the research into a structured format for the Slogan Generator.
- Sequence: Google Image Search > Keyword Extraction > Visual Inspiration Gathering > Research Compilation

F) SUBAGENT 2: Slogan Generator
- Name: Slogan Generator
- Final Output: A list of humorous slogans or puns for each niche.
- Tasks:
  1. Receive niche-specific research from the Niche Researcher.
  2. Use custom LLM prompts to generate multiple slogan ideas based on the research.
  3. Employ techniques like play-on-words and puns to create humorous slogans.
  4. Run the prompts multiple times to generate a diverse set of options.
- Sequence: Niche Research Input > LLM Prompt Generation > Slogan Generation > Slogan List Output

G) SUBAGENT 3: Slogan Selector
- Name: Slogan Selector
- Final Output: A curated list of the best slogans for each niche.
- Tasks:
  1. Receive the list of generated slogans from the Slogan Generator.
  2. Use an AI model to evaluate the humor and relevance of each slogan.
  3. Select the top slogans based on the AI model's scores.
  4. Compile the selected slogans into a curated list for each niche.
- Sequence: Slogan List Input > AI Evaluation > Top Slogan Selection > Curated Slogan List Output

H) SUBAGENT 4: Doodle Artist
- Name: Doodle Artist
- Final Output: AI-generated doodle-style images relevant to each niche.
- Tasks:
  1. Receive niche-specific information from the Niche Researcher.
  2. Use AI image generation tools like DALL-E to create doodle-style images.
  3. Ensure the images maintain a hand-drawn aesthetic and are relevant to the niche.
  4. Generate multiple images for each niche to provide variety.
- Sequence: Niche Information Input > AI Image Generation > Doodle Style Refinement > Doodle Image Output

I) SUBAGENT 5: Template Designer
- Name: Template Designer
- Final Output: 2-3 birthday card layout templates in Bannerbear, with placeholders for text and images.
- Tasks:
  1. Design visually appealing birthday card layouts in Bannerbear.
  2. Include specific areas for "card-text-line-1", "card-text-line-2", and "doodle-image-url".
  3. Ensure the templates are easily customizable and support 1000x1000 pixel images.
  4. Create multiple templates to provide variety in the final card designs.
- Sequence: Layout Design > Placeholder Integration > Template Creation in Bannerbear

J) SUBAGENT 6: Card Assembler
- Name: Card Assembler
- Final Output: A collection of funny birthday card images in JPEG format, sized at 1000x1000 pixels.
- Tasks:
  1. Receive the curated slogans from the Slogan Selector and doodle images from the Doodle Artist.
  2. Retrieve the birthday card templates from Bannerbear.
  3. Insert the slogans into the "card-text-line-1" and "card-text-line-2" placeholders in the templates.
  4. Add the doodle images to the "doodle-image-url" placeholder in the templates.
  5. Generate the final birthday card images for each niche using the populated templates.
  6. Save the images in JPEG format at 1000x1000 pixels.
- Sequence: Slogan and Doodle Image Input > Template Retrieval > Text and Image Insertion > Final Image Generation > JPEG Output

These subagents work together to create a streamlined workflow for generating funny, niche-specific birthday cards. The modular nature of the subagents allows for easy modification and improvement of individual components without disrupting the entire process.

subagent-summary-amended-for-actual-tasks..1

Here is the updated list of subagents and tasks to complete the birthday card generation workflow:

SUBAGENT 1: Niche Researcher
- Final Output: A list of relevant niches or topics for birthday cards, along with niche-specific keywords, phrases, and visual inspiration.
- Tasks:
  1. Search Google Images for "{niche topic} birthday cards" to identify popular niches. (Task 19)
  2. Use GPT-4 to analyze search results and extract relevant keywords and phrases for each niche. (Task 1)
  3. Save the extracted keywords and phrases to a text file on the server. (Task 9)
  4. Compile the research into a structured format for the Slogan Generator using GPT-4. (Task 1)

SUBAGENT 2: Slogan Generator
- Final Output: A list of humorous slogans or puns for each niche.
- Tasks:
  1. Access the niche-specific research from the text file saved by the Niche Researcher. (Task 10)
  2. Use the Prompt Generator to create custom LLM prompts for generating slogans based on the research. (Task 2)
  3. Run the generated prompts using GPT-4 to generate multiple slogan ideas for each niche. (Task 1)
  4. Save the generated slogans to a text file on the server. (Task 9)

SUBAGENT 3: Slogan Selector
- Final Output: A curated list of the best slogans for each niche.
- Tasks:
  1. Access the list of generated slogans from the text file saved by the Slogan Generator. (Task 10)
  2. Use GPT-4 to evaluate the humor and relevance of each slogan. (Task 1)
  3. Select the top slogans based on the GPT-4 evaluation.
  4. Save the selected slogans to a text file on the server. (Task 9)

SUBAGENT 4: Doodle Artist
- Final Output: AI-generated doodle-style images relevant to each niche.
- Tasks:
  1. Access the niche-specific keywords and phrases from the text file saved by the Niche Researcher. (Task 10)
  2. Use the Prompt Generator to create text prompts for generating doodle-style images using the keywords and phrases. (Task 2)
  3. Create square doodle-style images using Dalle3 based on the generated prompts. (Task 22)
  4. Save the generated images to the server. (Task 18)

SUBAGENT 5: Template Designer
- Final Output: 2-3 birthday card layout templates in Bannerbear, with placeholders for text and images.
- Tasks:
  1. Use Bannerbear to create 2-3 visually appealing birthday card templates with placeholders for text and images. (Task 26)
  2. Save the template URLs to a text file on the server. (Task 9)

SUBAGENT 6: Card Assembler
- Final Output: A collection of funny birthday card images in JPEG format, sized at 1000x1000 pixels.
- Tasks:
  1. Access the curated slogans from the text file saved by the Slogan Selector. (Task 10)
  2. Access the doodle images from the server saved by the Doodle Artist. (Task 11)
  3. Access the birthday card templates from the text file saved by the Template Designer. (Task 10)
  4. Use Bannerbear to insert the slogans and doodle images into the templates. (Task 26)
  5. Generate the final birthday card images using Bannerbear.
  6. Save the generated images to the server in JPEG format. (Task 18)

These subagents and tasks work together to create the automated workflow for generating funny, niche-specific birthday cards. The tasks available cover all the necessary functionalities, so no additional API calls or Python libraries are required.

subagent-summary-amended-for-actual-tasks..2

Thank you for providing the updated list of available tasks. Based on this information, I will revise the subagents and their specific tasks to complete the automated workflow for generating funny, niche-specific birthday card images.

A) FULL TASK SUMMARY: Create an automated workflow for generating funny, niche-specific birthday card images using AI-powered research, slogan generation, and image creation.

B) FINAL TASK OUTPUT: A collection of JPEG images, each sized at 1000x1000 pixels, featuring humorous slogans and AI-generated doodle-style images tailored to specific niches or topics.

C) COMPONENTS OVERVIEW:
1. Niche/Topic Research
2. Slogan Generation
3. Slogan Selection
4. Doodle Image Generation
5. Birthday Card Template Creation
6. Final Birthday Card Image Generation

E) SUBAGENT 1: Niche Researcher
- Name: Niche Researcher
- Final Output: A list of relevant niches or topics for birthday cards, along with niche-specific keywords, phrases, and visual inspiration.
- Tasks:
  1. Search Google (Task 15) for "{niche topic} birthday cards" to identify popular niches.
  2. GPT4 (Task 1) to analyze search results and extract relevant keywords and phrases for each niche.
  3. Search Google Images (Task 19) for top-selling cards in each niche to gather visual inspiration.
  4. GPT4 (Task 1) to compile the research into a structured format for the Slogan Generator.
- Sequence: Google Search > GPT4 Analysis > Google Image Search > GPT4 Compilation

F) SUBAGENT 2: Slogan Generator
- Name: Slogan Generator
- Final Output: A list of humorous slogans or puns for each niche.
- Tasks:
  1. GPT4 (Task 1) to receive niche-specific research from the Niche Researcher.
  2. Prompt Generator (Task 2) to create custom LLM prompts based on the research.
  3. GPT4 (Task 1) to generate multiple slogan ideas using the custom prompts, employing techniques like play-on-words and puns.
  4. GPT4 (Task 1) to compile the generated slogans into a list for each niche.
- Sequence: GPT4 Research Input > Prompt Generation > GPT4 Slogan Generation > GPT4 Slogan List Compilation

G) SUBAGENT 3: Slogan Selector
- Name: Slogan Selector
- Final Output: A curated list of the best slogans for each niche.
- Tasks:
  1. GPT4 (Task 1) to receive the list of generated slogans from the Slogan Generator.
  2. GPT4 (Task 1) to evaluate the humor and relevance of each slogan.
  3. GPT4 (Task 1) to select the top slogans based on the evaluation scores.
  4. GPT4 (Task 1) to compile the selected slogans into a curated list for each niche.
- Sequence: GPT4 Slogan List Input > GPT4 Evaluation > GPT4 Top Slogan Selection > GPT4 Curated Slogan List Output

H) SUBAGENT 4: Doodle Artist
- Name: Doodle Artist
- Final Output: AI-generated doodle-style images relevant to each niche.
- Tasks:
  1. GPT4 (Task 1) to receive niche-specific information from the Niche Researcher.
  2. Create square image with Dalle3 (Task 22) to generate doodle-style images based on the niche information.
  3. GPT4 (Task 1) to ensure the images maintain a hand-drawn aesthetic and are relevant to the niche.
  4. Resize image (Task 27) to scale the generated images to 1000x1000 pixels.
- Sequence: GPT4 Niche Information Input > Dalle3 Image Generation > GPT4 Doodle Style Verification > Image Resizing

I) SUBAGENT 5: Template Designer
- Name: Template Designer
- Final Output: 2-3 birthday card layout templates in Bannerbear, with placeholders for text and images.
- Tasks:
  1. GPT4 (Task 1) to design visually appealing birthday card layouts.
  2. Bannerbear - image creator (Task 26) to set up the templates with specific areas for "card-text-line-1", "card-text-line-2", and "doodle-image-url".
  3. GPT4 (Task 1) to ensure the templates are easily customizable and support 1000x1000 pixel images.
  4. Bannerbear - image creator (Task 26) to create multiple template variations for design variety.
- Sequence: GPT4 Layout Design > Bannerbear Template Setup > GPT4 Template Verification > Bannerbear Template Variation Creation

J) SUBAGENT 6: Card Assembler
- Name: Card Assembler
- Final Output: A collection of funny birthday card images in JPEG format, sized at 1000x1000 pixels.
- Tasks:
  1. GPT4 (Task 1) to receive the curated slogans from the Slogan Selector and doodle images from the Doodle Artist.
  2. Bannerbear - image creator (Task 26) to retrieve the birthday card templates.
  3. Bannerbear - image creator (Task 26) to insert the slogans and doodle images into the appropriate placeholders in the templates.
  4. Bannerbear - image creator (Task 26) to generate the final birthday card images for each niche using the populated templates.
  5. Save content to server (Task 9) to save the generated images in JPEG format at 1000x1000 pixels.
- Sequence: GPT4 Slogan and Doodle Image Input > Bannerbear Template Retrieval > Bannerbear Text and Image Insertion > Bannerbear Final Image Generation > JPEG Output to Server

This updated list of subagents and their specific tasks leverages the available tasks provided to create a streamlined workflow for generating funny, niche-specific birthday cards. The modular nature of the subagents allows for easy modification and improvement of individual components without disrupting the entire process.

subagent-summary-amended-for-actual-tasks GPT..3

Based on the available tasks, here is the updated list of subagents with specific tasks assigned to each component to achieve the final output:

**Subagent for Niche/Topic Research:**
- Name: Niche Researcher
- Final Output: A list of relevant niches or topics, including keywords and visual inspiration.
- Tasks:
  - Use Task 19 (Search Google images) with input "{niche topic} birthday cards" to collect image URLs for analysis and inspiration.
  - Use Task 5 (Extract visual data from image) on a selection of the fetched image URLs to gain insights into popular elements and styles within the niche.
  - Use Task 21 (Find related keywords) to obtain related keywords and phrases for the niche based on the initial input description.

**Subagent for Slogan Generation:**
- Name: Slogan Generator
- Final Output: A list of humorous slogans or puns for each niche.
- Tasks:
  - Use Task 2 (Prompt generator) to craft effective prompts for generating slogans tailored to the niche topics.
  - Use Task 1 (GPT4) with tailored prompts to generate multiple slogan ideas.
  - Repeat GPT4 calls as necessary to produce a diverse range of slogans and puns for each niche.

**Subagent for Slogan Selection:**
- Name: Slogan Selector
- Final Output: A curated list of the best slogans for each niche.
- Tasks:
  - Use Task 1 (GPT4) to evaluate the humor and relevance of each generated slogan.
  - Apply logic within GPT4 or Task 3 (Code interpreter) to score and rank the slogans, selecting the top choices for each niche.

**Subagent for Doodle Image Generation:**
- Name: Doodle Artist
- Final Output: AI-generated doodle-style images relevant to each niche.
- Tasks:
  - Use Task 22 (Create square image with Dalle3) for generating doodle-style images tailored to each slogan and niche.

**Subagent for Birthday Card Template Creation:**
- Name: Template Designer
- Final Output: Birthday card layout templates ready for content insertion.
- Tasks:
  - REQUIRED: Define a custom task to interact with Bannerbear directly to create and manipulate templates, or use an API call as a placeholder until such functionality is integrated into the list of available tasks.

**Subagent for Final Birthday Card Image Generation:**
- Name: Card Assembler
- Final Output: A collection of funny birthday card images in JPEG format, sized at 1000x1000 pixels.
- Tasks:
  - Use Task 26 (Bannerbear - image creator) to feed in selected slogans and doodle images into predefined Bannerbear templates, create the cards, and output them as JPEG images.

**Additional Processes:**
- Scaling up designs: Use Task 29 (Upscale image) to scale up the final designs if necessary.
- Server management: Throughout the process, Tasks 8 (Move files around server), 9 (Save content to server), and 12 (Find/replace characters in file(s)) may be used for intermediate file management to streamline the workflow.

**Custom API or Python Library (REQUIRED):**
- For the direct creation and editing of Bannerbear templates, you will need a custom task that allows for setting up templates and feeding data into these templates. This could take the form of a dedicated API call or Python library that interacts with Bannerbear's API, allowing the subagent to define placeholders, insert images and text, and generate the final birthday card layouts.

subagent-summary-amended-for-actual-tasks

A) FULL TASK SUMMARY: Create an automated workflow for generating funny, niche-specific birthday card images using AI-powered research, slogan generation, and image creation.

B) FINAL TASK OUTPUT: A collection of JPEG images, each sized at 1000x1000 pixels, featuring humorous slogans and AI-generated doodle-style images tailored to specific niches or topics.

C) COMPONENTS OVERVIEW:
1. Niche/Topic Research
2. Slogan Generation
3. Slogan Selection
4. Doodle Image Generation
5. Birthday Card Template Creation
6. Final Birthday Card Image Generation

E) SUBAGENT 1: Niche Researcher
- Name: Niche Researcher
- Final Output: A list of relevant niches or topics for birthday cards, along with niche-specific keywords, phrases, and visual inspiration.
- Tasks:
  1. Search Google Images (Task 19) for "{niche topic} birthday cards" to identify popular niches and gather visual inspiration.
  2. Extract visual data from image (Task 5) on a selection of the fetched image URLs to gain insights into popular elements and styles within the niche.
  3. Find related keywords (Task 21) to obtain related keywords and phrases for the niche based on the initial input description.
  4. GPT4 (Task 1) to analyze search results and extracted data, then compile the research into a structured format for the Slogan Generator.
- Sequence: Google Image Search > Image Data Extraction > Related Keyword Search > GPT4 Analysis and Compilation

F) SUBAGENT 2: Slogan Generator
- Name: Slogan Generator
- Final Output: A list of humorous slogans or puns for each niche.
- Tasks:
  1. Prompt Generator (Task 2) to create custom LLM prompts based on the niche-specific research from the Niche Researcher.
  2. GPT4 (Task 1) to generate multiple slogan ideas using the custom prompts, employing techniques like play-on-words and puns.
  3. GPT4 (Task 1) to compile the generated slogans into a list for each niche.
- Sequence: Prompt Generation > GPT4 Slogan Generation > GPT4 Slogan List Compilation

G) SUBAGENT 3: Slogan Selector
- Name: Slogan Selector
- Final Output: A curated list of the best slogans for each niche.
- Tasks:
  1. GPT4 (Task 1) to evaluate the humor and relevance of each slogan generated by the Slogan Generator.
  2. Code Interpreter (Task 3) to score and rank the slogans, selecting the top choices for each niche.
  3. GPT4 (Task 1) to compile the selected slogans into a curated list for each niche.
- Sequence: GPT4 Evaluation > Code Interpreter Scoring and Ranking > GPT4 Curated List Compilation

H) SUBAGENT 4: Doodle Artist
- Name: Doodle Artist
- Final Output: AI-generated doodle-style images relevant to each niche.
- Tasks:
  1. GPT4 (Task 1) to receive niche-specific information from the Niche Researcher and generate prompts for doodle-style images.
  2. Create square image with Dalle3 (Task 22) to generate doodle-style images based on the prompts.
  3. Resize image (Task 27) to scale the generated images to 1000x1000 pixels.
- Sequence: GPT4 Prompt Generation > Dalle3 Image Generation > Image Resizing

I) SUBAGENT 5: Template Designer
- Name: Template Designer
- Final Output: 2-3 birthday card layout templates in Bannerbear, with placeholders for text and images.
- Tasks:
  1. GPT4 (Task 1) to design visually appealing birthday card layouts and provide instructions for creating templates in Bannerbear.
  2. REQUIRED: Custom API or Python Library to interact with Bannerbear directly for creating and editing templates, setting up placeholders for text and images.
  3. GPT4 (Task 1) to ensure the templates are easily customizable and support 1000x1000 pixel images.
- Sequence: GPT4 Layout Design > Custom Bannerbear Integration > GPT4 Template Verification

J) SUBAGENT 6: Card Assembler
- Name: Card Assembler
- Final Output: A collection of funny birthday card images in JPEG format, sized at 1000x1000 pixels.
- Tasks:
  1. GPT4 (Task 1) to receive the curated slogans from the Slogan Selector and doodle images from the Doodle Artist.
  2. REQUIRED: Custom API or Python Library to retrieve the birthday card templates from Bannerbear and insert the slogans and doodle images into the appropriate placeholders.
  3. Save content to server (Task 9) to save the generated images in JPEG format at 1000x1000 pixels.
- Sequence: GPT4 Slogan and Doodle Image Input > Custom Bannerbear Integration for Template Population > JPEG Output to Server

This automated workflow leverages the available tasks and incorporates a custom API or Python library for seamless integration with Bannerbear. The modular subagents allow for efficient generation of humorous, niche-specific birthday cards, while the custom integration ensures smooth template creation and population. The output is a collection of high-quality, funny birthday card images tailored to various niches and topics.

random-autocomplete

To achieve the end goal of creating a collection of funny birthday card images tailored to specific niches, the AI Agent could follow this sequence of tasks:

1. Research: Conduct a Google image search for "{niche topic} birthday cards" to gather information on current top-selling images in the niche and identify relevant themes, styles, and elements.

2. Niche Selection: Based on the research, select a few relevant niches or topics to focus on for the birthday card generation.

3. Slogan Generation:
   a. Perform a series of searches to collect possible words, phrases, and puns relevant to each selected niche.
   b. Use natural language processing (NLP) techniques to analyze the search results and generate a summary of key information.
   c. Develop custom prompts that utilize the summarized research to generate a list of humorous slogans and puns for each niche.
   d. Run these prompts multiple times to create a diverse set of slogan options.
   e. Employ another AI model to review and select the best slogans for each niche.

4. Image Generation:
   a. Use an AI image generator (e.g., DALL-E) to create doodle-style images that visually represent each niche and complement the chosen slogans.
   b. Ensure the generated images have a hand-drawn appearance to match the desired style.

5. Template Creation:
   a. Use the Bannerbear tool to create 2-3 template image structures for the birthday cards.
   b. Design the templates to accommodate the selected slogans and generated images.

6. Card Assembly:
   a. Integrate the chosen slogans and generated images into the Bannerbear templates.
   b. Use placeholders like "card-text-line-1", "card-text-line-2", and "doodle-image-url" to ensure proper placement of the content within the templates.

7. Output Generation:
   a. Render the final birthday card images using the Bannerbear tool, generating 1000x1000 pixel JPEG files.
   b. Ensure the images are easily scalable for future use.

The overall sequence would be:
Research > Niche Selection > Slogan Generation > Image Generation > Template Creation > Card Assembly > Output Generation

By following this sequence of tasks, the AI Agent can leverage its access to various APIs, libraries, and AI models to generate a collection of humorous, niche-specific birthday card images that meet the desired end goal.

claude 3 - mermaid-for-all

<!DOCTYPE html>
<html>
<head>
  <title>Mermaid Diagram Example</title>
  <script src="https://cdnjs.cloudflare.com/ajax/libs/mermaid/9.4.3/mermaid.min.js"></script>
</head>
<body>
  <div class="mermaid">
graph TD
A[Niche/Topic Description Input] --> B{Niche Researcher}
B --> C[Google Image Search for niche birthday cards]
C --> D[Extract visual data from images]
C --> E[Find related keywords]
D --> F[GPT4 analysis and compilation]
E --> F
F --> G{Slogan Generator}
G --> H[Prompt Generator creates custom LLM prompts]
H --> I[GPT4 generates multiple slogan ideas]
I --> J[GPT4 compiles generated slogans into a list]
J --> K{Slogan Selector}
K --> L[GPT4 evaluates humor and relevance of slogans]
L --> M[Code Interpreter scores and ranks slogans]
M --> N[GPT4 compiles selected slogans into a curated list]
B --> O{Doodle Artist}
O --> P[GPT4 generates prompts for doodle-style images]
P --> Q[Create square images with Dalle3]
Q --> R[Resize images to 1000x1000 pixels]
S{Template Designer} --> T[GPT4 designs birthday card layouts]
T --> U[Custom Bannerbear Integration for template creation]
U --> V[GPT4 verifies template customizability]
N --> W{Card Assembler}
R --> W
V --> W
W --> X[GPT4 receives curated slogans and doodle images]
X --> Y[Custom Bannerbear Integration for template population]
Y --> Z[Save generated images as 1000x1000 JPEG on server]
  </div>

SubAgent #1

claude 3 - subagent1

subagent1-refined

A) SUBAGENT SUMMARY: The Niche Researcher subagent conducts research to identify relevant niches and gather niche-specific information for generating birthday cards.

B) FINAL TASK OUTPUT: A structured JSON file containing a list of relevant niches or topics for birthday cards, along with niche-specific keywords, phrases, and visual inspiration. The JSON file will be saved on the server with a URL for easy access by other subagents.

C) SUBAGENT INPUT: A text prompt (1-500 characters) describing the general theme or audience for the birthday cards (e.g., "funny birthday cards for dog lovers" or "heartwarming birthday cards for grandmothers").

E) SUBAGENT TASK SUMMARY:
1. GPT4 (Task 1) receives the initial text prompt and generates a list of 5-10 potential niches or topics related to the theme or audience.
2. For each niche or topic:
   a. Search Google Images (Task 19) for "{niche topic} birthday cards" to identify popular visual styles and gather image URLs.
   b. GPT4 (Task 1) selects the top 5 most relevant and visually inspiring image URLs from the search results.
   c. For each selected image URL:
      i. Save external web file to server (Task 18) to store the image on the local server.
      ii. Extract visual data from image (Task 5) to gain insights into popular elements and styles within the niche.
   d. Find related keywords (Task 21) to obtain related keywords and phrases for the niche based on the initial niche description.
   e. GPT4 (Task 1) analyzes the extracted visual data and related keywords, then compiles the research into a structured JSON format for the niche.
3. GPT4 (Task 1) combines the JSON data for all niches into a single structured JSON file.
4. Save content to server (Task 9) to store the final JSON file on the server with a unique URL.

F) SILOS:
Silo 1: Niche Identification
- GPT4 (Task 1) receives the initial text prompt and generates a list of 5-10 potential niches or topics related to the theme or audience.

Silo 2: Visual Inspiration Research (for each niche)
- Search Google Images (Task 19) for "{niche topic} birthday cards" to identify popular visual styles and gather image URLs.
- GPT4 (Task 1) selects the top 5 most relevant and visually inspiring image URLs from the search results.
- For each selected image URL:
  - Save external web file to server (Task 18) to store the image on the local server.
  - Extract visual data from image (Task 5) to gain insights into popular elements and styles within the niche.

Silo 3: Keyword Research (for each niche)
- Find related keywords (Task 21) to obtain related keywords and phrases for the niche based on the initial niche description.

Silo 4: Data Compilation and Storage
- GPT4 (Task 1) analyzes the extracted visual data and related keywords, then compiles the research into a structured JSON format for each niche.
- GPT4 (Task 1) combines the JSON data for all niches into a single structured JSON file.
- Save content to server (Task 9) to store the final JSON file on the server with a unique URL.

subagent1refined-repeating

G) FULL TASKS (INCLUDING REPEATING TASKS):
1. GPT4 (Task 1) receives the initial text prompt and generates a list of 5-10 potential niches or topics related to the theme or audience.
2. For each niche or topic (assuming 5-10 niches):
   a. Search Google Images (Task 19) for "{niche topic} birthday cards" to identify popular visual styles and gather image URLs.
   b. GPT4 (Task 1) selects the top 5 most relevant and visually inspiring image URLs from the search results.
   c. For each selected image URL (assuming 5 images per niche):
      i. Save external web file to server (Task 18) to store the image on the local server.
      ii. Extract visual data from image (Task 5) to gain insights into popular elements and styles within the niche.
   d. Find related keywords (Task 21) to obtain related keywords and phrases for the niche based on the initial niche description.
   e. GPT4 (Task 1) analyzes the extracted visual data and related keywords, then compiles the research into a structured JSON format for the niche.
3. GPT4 (Task 1) combines the JSON data for all niches into a single structured JSON file.
4. Save content to server (Task 9) to store the final JSON file on the server with a unique URL.

Expanded task list:
1. GPT4 (Task 1) - Generate list of niches
2. Search Google Images (Task 19) - Niche 1
3. GPT4 (Task 1) - Select top 5 images for Niche 1
4. Save external web file to server (Task 18) - Image 1, Niche 1
5. Extract visual data from image (Task 5) - Image 1, Niche 1
6. Save external web file to server (Task 18) - Image 2, Niche 1
7. Extract visual data from image (Task 5) - Image 2, Niche 1
8. Save external web file to server (Task 18) - Image 3, Niche 1
9. Extract visual data from image (Task 5) - Image 3, Niche 1
10. Save external web file to server (Task 18) - Image 4, Niche 1
11. Extract visual data from image (Task 5) - Image 4, Niche 1
12. Save external web file to server (Task 18) - Image 5, Niche 1
13. Extract visual data from image (Task 5) - Image 5, Niche 1
14. Find related keywords (Task 21) - Niche 1
15. GPT4 (Task 1) - Compile research for Niche 1
16. Search Google Images (Task 19) - Niche 2
...
(Repeat tasks 3-15 for each additional niche, assuming 5-10 niches in total)
...
N-3. Find related keywords (Task 21) - Niche 10
N-2. GPT4 (Task 1) - Compile research for Niche 10
N-1. GPT4 (Task 1) - Combine JSON data for all niches
N. Save content to server (Task 9) - Final JSON file

Subagent1 task-ID-numbers

subagent1-output-variables

Here's the updated task list with the output tokens added:

G) FULL TASKS (INCLUDING REPEATING TASKS):
1. GPT4 (Task 1) receives the initial text prompt and generates a list of 5-10 potential niches or topics related to the theme or audience. [niches]
2. For each niche or topic (assuming 5-10 niches):
   a. Search Google Images (Task 19) for "{niche topic} birthday cards" to identify popular visual styles and gather image URLs. [niche-image-urls]
   b. GPT4 (Task 1) selects the top 5 most relevant and visually inspiring image URLs from the search results. [niche-top-image-urls]
   c. For each selected image URL (assuming 5 images per niche):
      i. Save external web file to server (Task 18) to store the image on the local server. [niche-image-file]
      ii. Extract visual data from image (Task 5) to gain insights into popular elements and styles within the niche. [niche-image-data]
   d. Find related keywords (Task 21) to obtain related keywords and phrases for the niche based on the initial niche description. [niche-keywords]
   e. GPT4 (Task 1) analyzes the extracted visual data and related keywords, then compiles the research into a structured JSON format for the niche. [niche-research-json]
3. GPT4 (Task 1) combines the JSON data for all niches into a single structured JSON file. [combined-research-json]
4. Save content to server (Task 9) to store the final JSON file on the server with a unique URL. [final-json-url]

H) OUTPUT TOKENS

Expanded task list:
1. GPT4 (Task 1) - Generate list of niches [niches]
2. Search Google Images (Task 19) - Niche 1 [niche1-image-urls]
3. GPT4 (Task 1) - Select top 5 images for Niche 1 [niche1-top-image-urls]
4. Save external web file to server (Task 18) - Image 1, Niche 1 [niche1-image1-file]
5. Extract visual data from image (Task 5) - Image 1, Niche 1 [niche1-image1-data]
6. Save external web file to server (Task 18) - Image 2, Niche 1 [niche1-image2-file]
7. Extract visual data from image (Task 5) - Image 2, Niche 1 [niche1-image2-data]
8. Save external web file to server (Task 18) - Image 3, Niche 1 [niche1-image3-file]
9. Extract visual data from image (Task 5) - Image 3, Niche 1 [niche1-image3-data]
10. Save external web file to server (Task 18) - Image 4, Niche 1 [niche1-image4-file]
11. Extract visual data from image (Task 5) - Image 4, Niche 1 [niche1-image4-data]
12. Save external web file to server (Task 18) - Image 5, Niche 1 [niche1-image5-file]
13. Extract visual data from image (Task 5) - Image 5, Niche 1 [niche1-image5-data]
14. Find related keywords (Task 21) - Niche 1 [niche1-keywords]
15. GPT4 (Task 1) - Compile research for Niche 1 [niche1-research-json]
16. Search Google Images (Task 19) - Niche 2 [niche2-image-urls]
...
(Repeat tasks 3-15 for each additional niche, assuming 5-10 niches in total)
...
N-3. Find related keywords (Task 21) - Niche 10 [niche10-keywords]
N-2. GPT4 (Task 1) - Compile research for Niche 10 [niche10-research-json]
N-1. GPT4 (Task 1) - Combine JSON data for all niches [combined-research-json]
N. Save content to server (Task 9) - Final JSON file [final-json-url]

subagent1refined-repeating-json-core

A) SUBAGENT SUMMARY: Research and compile visual and keyword data for multiple niches related to a given theme or audience for generating birthday card ideas.

B) FINAL TASK OUTPUT: A structured JSON file containing compiled research data for 5-10 niches, including top image URLs, extracted visual elements, and related keywords, stored on the server with a unique URL.

C) SUBAGENT INPUT:
- Initial text prompt describing the theme or audience for birthday card ideas

E) SUBAGENT TASK SUMMARY:
Initial text prompt > Generate list of 5-10 niches > For each niche: Search Google Images, select top 5 images, save images to server, extract visual data from images, find related keywords > Compile research for each niche > Combine JSON data for all niches > Save final JSON file to server

F) SILOS: N/A

G) TASKFLOW SUMMARY:

Task 1:
Title task: Generate list of niches
Description of the task: GPT4 generates a list of 5-10 potential niches or topics related to the theme or audience based on the initial text prompt.
Input(s) involved: Initial text prompt
Output token name: List of niches

Task 2:
Title task: Search Google Images for niche
Description of the task: Search Google Images for "{niche topic} birthday cards" to identify popular visual styles and gather image URLs for each niche.
Input(s) involved: Niche topic
Output token name: Image search results

Task 3:
Title task: Select top 5 images for niche
Description of the task: GPT4 selects the top 5 most relevant and visually inspiring image URLs from the search results for each niche.
Input(s) involved: Image search results
Output token name: Top 5 image URLs

Task 4:
Title task: Save image to server
Description of the task: Save each selected image URL to the local server using the "Save external web file to server" task.
Input(s) involved: Image URL
Output token name: Saved image file

Task 5:
Title task: Extract visual data from image
Description of the task: Use the "Extract visual data from image" task to gain insights into popular elements and styles within each saved image.
Input(s) involved: Saved image file
Output token name: Extracted visual data

Task 6:
Title task: Find related keywords for niche
Description of the task: Use the "Find related keywords" task to obtain related keywords and phrases for each niche based on the initial niche description.
Input(s) involved: Niche topic
Output token name: Related keywords

Task 7:
Title task: Compile research for niche
Description of the task: GPT4 analyzes the extracted visual data and related keywords, then compiles the research into a structured JSON format for each niche.
Input(s) involved: Extracted visual data, Related keywords
Output token name: Niche research JSON

Task 8:
Title task: Combine JSON data for all niches
Description of the task: GPT4 combines the JSON data for all niches into a single structured JSON file.
Input(s) involved: Niche research JSON (for each niche)
Output token name: Combined research JSON

Task 9:
Title task: Save final JSON file to server
Description of the task: Use the "Save content to server" task to store the final combined JSON file on the server with a unique URL.
Input(s) involved: Combined research JSON
Output token name: Final JSON file URL

subagent1mermaid

SubAgent #2

claude 3 - subagent2

subagent2-refined

A) SUBAGENT SUMMARY: Slogan Generator is responsible for generating humorous slogans and puns based on niche-specific research.

B) FINAL TASK OUTPUT: A JSON file containing a list of 10-20 humorous slogans or puns for each niche, with each slogan being a maximum of 100 characters in length.

C) SUBAGENT INPUT: A JSON file containing niche-specific research data, including relevant keywords, phrases, and visual inspiration from top-selling cards in the niche.

D) SUBAGENT TASK SUMMARY:
1. Input: JSON file with niche-specific research data
2. Prompt Generator (Task 2) to create custom LLM prompts based on the niche-specific research data
   - Input: JSON file with niche-specific research data
   - Output: 3-5 custom LLM prompts tailored to the niche
3. GPT4 (Task 1) to generate multiple slogan ideas using the custom prompts
   - Input: Custom LLM prompts from step 2
   - Output: 20-30 slogan ideas per niche, employing techniques like play-on-words and puns
4. GPT4 (Task 1) to filter and refine the generated slogans
   - Input: Generated slogan ideas from step 3
   - Output: Top 10-20 slogans per niche, each with a maximum length of 100 characters
5. GPT4 (Task 1) to compile the refined slogans into a structured JSON format
   - Input: Top 10-20 slogans per niche from step 4
   - Output: JSON file containing the slogans organized by niche
6. Output: JSON file with 10-20 humorous slogans or puns for each niche

E) SILOS:
Silo 1: Prompt Generation
- Steps: 2
- Input: JSON file with niche-specific research data
- Output: 3-5 custom LLM prompts tailored to the niche

Silo 2: Slogan Generation and Refinement
- Steps: 3, 4
- Input: Custom LLM prompts from Silo 1
- Output: Top 10-20 slogans per niche, each with a maximum length of 100 characters

Silo 3: JSON Compilation
- Steps: 5
- Input: Top 10-20 slogans per niche from Silo 2
- Output: JSON file containing the slogans organized by niche

subagent2refined-repeating

G) FULL TASKS (INCLUDING REPEATING TASKS):

1. Input: JSON file with niche-specific research data
2. Prompt Generator (Task 2) to create custom LLM prompts based on the niche-specific research data
   - Input: JSON file with niche-specific research data
   - Output: 3-5 custom LLM prompts tailored to the niche
3. GPT4 (Task 1) to generate multiple slogan ideas using the custom prompts (Repeat steps 3a-3c for each custom prompt generated in step 2)
   a. Input: Custom LLM prompt
   b. Output: 20-30 slogan ideas employing techniques like play-on-words and puns
   c. Append the generated slogan ideas to a list
4. GPT4 (Task 1) to filter and refine the generated slogans
   - Input: List of all generated slogan ideas from step 3
   - Output: Top 10-20 slogans per niche, each with a maximum length of 100 characters
5. GPT4 (Task 1) to compile the refined slogans into a structured JSON format
   - Input: Top 10-20 slogans per niche from step 4
   - Output: JSON file containing the slogans organized by niche
6. Output: JSON file with 10-20 humorous slogans or puns for each niche

Subagent2 task-ID-numbers

subagent2-output-variables

Here's the updated task list with output tokens:

--------------
G) FULL TASKS (INCLUDING REPEATING TASKS):

1. Input: JSON file with niche-specific research data [niche-research-data]
2. Prompt Generator (Task 2) to create custom LLM prompts based on the niche-specific research data
   - Input: JSON file with niche-specific research data [niche-research-data]
   - Output: 3-5 custom LLM prompts tailored to the niche [custom-prompt1], [custom-prompt2], [custom-prompt3], [custom-prompt4], [custom-prompt5]
3. GPT4 (Task 1) to generate multiple slogan ideas using the custom prompts (Repeat steps 3a-3c for each custom prompt generated in step 2)
   a. Input: Custom LLM prompt [custom-prompt1], [custom-prompt2], [custom-prompt3], [custom-prompt4], [custom-prompt5]
   b. Output: 20-30 slogan ideas employing techniques like play-on-words and puns [slogan-ideas1], [slogan-ideas2], [slogan-ideas3], [slogan-ideas4], [slogan-ideas5]
   c. Append the generated slogan ideas to a list [all-slogan-ideas]
4. GPT4 (Task 1) to filter and refine the generated slogans
   - Input: List of all generated slogan ideas from step 3 [all-slogan-ideas]
   - Output: Top 10-20 slogans per niche, each with a maximum length of 100 characters [refined-slogans]
5. GPT4 (Task 1) to compile the refined slogans into a structured JSON format
   - Input: Top 10-20 slogans per niche from step 4 [refined-slogans]
   - Output: JSON file containing the slogans organized by niche [slogans-json]
6. Output: JSON file with 10-20 humorous slogans or puns for each niche [slogans-json]

H) OUTPUT TOKENS

[niche-research-data]: JSON file with niche-specific research data
[custom-prompt1]: Custom LLM prompt tailored to the niche
[custom-prompt2]: Custom LLM prompt tailored to the niche
[custom-prompt3]: Custom LLM prompt tailored to the niche
[custom-prompt4]: Custom LLM prompt tailored to the niche
[custom-prompt5]: Custom LLM prompt tailored to the niche
[slogan-ideas1]: 20-30 slogan ideas generated using [custom-prompt1]
[slogan-ideas2]: 20-30 slogan ideas generated using [custom-prompt2]
[slogan-ideas3]: 20-30 slogan ideas generated using [custom-prompt3]
[slogan-ideas4]: 20-30 slogan ideas generated using [custom-prompt4]
[slogan-ideas5]: 20-30 slogan ideas generated using [custom-prompt5]
[all-slogan-ideas]: List of all slogan ideas generated in step 3
[refined-slogans]: Top 10-20 slogans per niche, each with a maximum length of 100 characters
[slogans-json]: JSON file containing the slogans organized by niche
--------------

subagent2refined-repeating-json-core

A) SUBAGENT SUMMARY: A subagent that generates humorous slogans or puns for specific niches using custom LLM prompts and niche-specific research data.

B) FINAL TASK OUTPUT: A JSON file containing 10-20 humorous slogans or puns for each niche, with each slogan having a maximum length of 100 characters.

C) SUBAGENT INPUT: A JSON file with niche-specific research data.

E) SUBAGENT TASK SUMMARY:
1. Input: JSON file with niche-specific research data
2. Prompt Generator (Task 2) to create custom LLM prompts based on the niche-specific research data
3. GPT4 (Task 1) to generate multiple slogan ideas using the custom prompts (Repeat for each custom prompt)
4. GPT4 (Task 1) to filter and refine the generated slogans
5. GPT4 (Task 1) to compile the refined slogans into a structured JSON format
6. Output: JSON file with 10-20 humorous slogans or puns for each niche

F) SILOS: N/A

G) TASKFLOW SUMMARY:

Task 1: Generate custom LLM prompts
Description of the task: Use the Prompt Generator (Task 2) to create 3-5 custom LLM prompts tailored to the niche based on the niche-specific research data.
Input(s) involved: JSON file with niche-specific research data
Output token name: Custom LLM prompts

Task 2: Generate slogan ideas using custom prompts
Description of the task: Use GPT4 (Task 1) to generate 20-30 slogan ideas for each custom prompt, employing techniques like play-on-words and puns. Append the generated slogan ideas to a list.
Input(s) involved: Custom LLM prompts
Output token name: List of generated slogan ideas

Task 3: Filter and refine generated slogans
Description of the task: Use GPT4 (Task 1) to filter and refine the generated slogans, selecting the top 10-20 slogans per niche, each with a maximum length of 100 characters.
Input(s) involved: List of generated slogan ideas
Output token name: Top 10-20 slogans per niche

Task 4: Compile refined slogans into JSON format
Description of the task: Use GPT4 (Task 1) to compile the refined slogans into a structured JSON format, organizing the slogans by niche.
Input(s) involved: Top 10-20 slogans per niche
Output token name: JSON file with humorous slogans organized by niche

subagent2mermaid

SubAgent #3

claude 3 - subagent3

subagent1-refined

A) SUBAGENT SUMMARY: The Slogan Selector subagent evaluates and selects the best slogans generated by the Slogan Generator for each niche.

B) FINAL TASK OUTPUT: A curated list of the best slogans for each niche, compiled into a structured text format (e.g., JSON or CSV) containing the selected slogans, their corresponding niche, and relevance scores.

C) SUBAGENT INPUT:
- A list of generated slogans for each niche, including the slogans and their corresponding niche (text format, e.g., JSON or CSV)
- Niche-specific research data from the Niche Researcher, including relevant keywords, phrases, and visual inspiration (text format, e.g., JSON or CSV)

E) SUBAGENT TASK SUMMARY:
1. GPT4 (Task 1) receives the list of generated slogans and niche-specific research data as input.
2. GPT4 evaluates the humor and relevance of each slogan based on the niche-specific research, assigning a humor score (1-5) and a relevance score (1-5) to each slogan.
3. GPT4 outputs the slogans with their assigned scores in a structured format (e.g., JSON or CSV).
4. Code Interpreter (Task 3) receives the structured data containing slogans and their scores.
5. Code Interpreter sorts and ranks the slogans based on their combined humor and relevance scores, selecting the top 5-10 slogans for each niche.
6. Code Interpreter outputs the selected slogans for each niche in a structured format (e.g., JSON or CSV).
7. GPT4 (Task 1) receives the selected slogans for each niche.
8. GPT4 compiles the selected slogans into a curated list for each niche, ensuring the list is well-structured and easily readable.
9. GPT4 outputs the final curated list of the best slogans for each niche in a structured text format (e.g., JSON or CSV).

F) SILOS:
Silo 1: Slogan Evaluation
- GPT4 receives the list of generated slogans and niche-specific research data.
- GPT4 evaluates the humor and relevance of each slogan, assigning scores.
- GPT4 outputs the slogans with their assigned scores in a structured format.

Silo 2: Slogan Ranking and Selection
- Code Interpreter receives the structured data containing slogans and their scores.
- Code Interpreter sorts and ranks the slogans based on their combined scores.
- Code Interpreter selects the top 5-10 slogans for each niche.
- Code Interpreter outputs the selected slogans for each niche in a structured format.

Silo 3: Curated List Compilation
- GPT4 receives the selected slogans for each niche.
- GPT4 compiles the selected slogans into a curated list for each niche.
- GPT4 ensures the list is well-structured and easily readable.
- GPT4 outputs the final curated list of the best slogans for each niche in a structured text format.

subagent1refined-repeating

A) SUBAGENT SUMMARY: The Slogan Selector subagent evaluates and selects the best slogans generated by the Slogan Generator for each niche.

G) FULL TASKS (INCLUDING REPEATING TASKS):
1. GPT4 (Task 1) receives the list of generated slogans and niche-specific research data as input.
2. For each niche:
   a. For each slogan in the niche:
      i. GPT4 evaluates the humor of the slogan based on the niche-specific research, assigning a humor score (1-5).
      ii. GPT4 evaluates the relevance of the slogan based on the niche-specific research, assigning a relevance score (1-5).
3. GPT4 outputs the slogans with their assigned scores in a structured format (e.g., JSON or CSV).
4. Code Interpreter (Task 3) receives the structured data containing slogans and their scores.
5. For each niche:
   a. Code Interpreter sorts and ranks the slogans based on their combined humor and relevance scores.
   b. Code Interpreter selects the top 5-10 slogans for the niche.
6. Code Interpreter outputs the selected slogans for each niche in a structured format (e.g., JSON or CSV).
7. GPT4 (Task 1) receives the selected slogans for each niche.
8. For each niche:
   a. GPT4 compiles the selected slogans into a curated list for the niche, ensuring the list is well-structured and easily readable.
9. GPT4 outputs the final curated list of the best slogans for each niche in a structured text format (e.g., JSON or CSV).

Subagent1 task-ID-numbers3

subagent3-output-variables

Here's the updated task list with the output tokens added:

A) SUBAGENT SUMMARY: The Slogan Selector subagent evaluates and selects the best slogans generated by the Slogan Generator for each niche.

C) SUBAGENT INPUT:
- A list of generated slogans for each niche, including the slogans and their corresponding niche (text format, e.g., JSON or CSV) [generated-slogans-list]
- Niche-specific research data from the Niche Researcher, including relevant keywords, phrases, and visual inspiration (text format, e.g., JSON or CSV) [niche-research-data]

E) SUBAGENT TASK SUMMARY:
1. GPT4 (Task 1) receives the list of generated slogans [generated-slogans-list] and niche-specific research data [niche-research-data] as input.
2. GPT4 evaluates the humor and relevance of each slogan based on the niche-specific research, assigning a humor score (1-5) and a relevance score (1-5) to each slogan.
3. GPT4 outputs the slogans with their assigned scores in a structured format (e.g., JSON or CSV). [evaluated-slogans-list]
4. Code Interpreter (Task 3) receives the structured data containing slogans and their scores [evaluated-slogans-list].
5. Code Interpreter sorts and ranks the slogans based on their combined humor and relevance scores, selecting the top 5-10 slogans for each niche.
6. Code Interpreter outputs the selected slogans for each niche in a structured format (e.g., JSON or CSV). [selected-slogans-list]
7. GPT4 (Task 1) receives the selected slogans for each niche [selected-slogans-list].
8. GPT4 compiles the selected slogans into a curated list for each niche, ensuring the list is well-structured and easily readable.
9. GPT4 outputs the final curated list of the best slogans for each niche in a structured text format (e.g., JSON or CSV). [best-slogans-list]

F) SILOS:
Silo 1: Slogan Evaluation
- GPT4 receives the list of generated slogans [generated-slogans-list] and niche-specific research data [niche-research-data].
- GPT4 evaluates the humor and relevance of each slogan, assigning scores.
- GPT4 outputs the slogans with their assigned scores in a structured format. [evaluated-slogans-list]

Silo 2: Slogan Ranking and Selection
- Code Interpreter receives the structured data containing slogans and their scores [evaluated-slogans-list].
- Code Interpreter sorts and ranks the slogans based on their combined scores.
- Code Interpreter selects the top 5-10 slogans for each niche.
- Code Interpreter outputs the selected slogans for each niche in a structured format. [selected-slogans-list]

Silo 3: Curated List Compilation
- GPT4 receives the selected slogans for each niche [selected-slogans-list].
- GPT4 compiles the selected slogans into a curated list for each niche.
- GPT4 ensures the list is well-structured and easily readable.
- GPT4 outputs the final curated list of the best slogans for each niche in a structured text format. [best-slogans-list]

G) FULL TASKS (INCLUDING REPEATING TASKS):
1. GPT4 (Task 1) receives the list of generated slogans [generated-slogans-list] and niche-specific research data [niche-research-data] as input.
2. For each niche [niche]:
   a. For each slogan [slogan] in the niche:
      i. GPT4 evaluates the humor of the slogan based on the niche-specific research, assigning a humor score (1-5). [slogan-humor-score]
      ii. GPT4 evaluates the relevance of the slogan based on the niche-specific research, assigning a relevance score (1-5). [slogan-relevance-score]
3. GPT4 outputs the slogans with their assigned scores in a structured format (e.g., JSON or CSV). [evaluated-slogans-list]
4. Code Interpreter (Task 3) receives the structured data containing slogans and their scores [evaluated-slogans-list].
5. For each niche [niche]:
   a. Code Interpreter sorts and ranks the slogans based on their combined humor and relevance scores. [sorted-slogans-list]
   b. Code Interpreter selects the top 5-10 slogans for the niche. [selected-slogans-list-niche]
6. Code Interpreter outputs the selected slogans for each niche in a structured format (e.g., JSON or CSV). [selected-slogans-list]
7. GPT4 (Task 1) receives the selected slogans for each niche [selected-slogans-list].
8. For each niche [niche]:
   a. GPT4 compiles the selected slogans into a curated list for the niche, ensuring the list is well-structured and easily readable. [curated-slogans-list-niche]
9. GPT4 outputs the final curated list of the best slogans for each niche in a structured text format (e.g., JSON or CSV). [best-slogans-list]

H) OUTPUT TOKENS:
- [best-slogans-list]: The final curated list of the best slogans for each niche in a structured text format.
- [generated-slogans-list]: A list of generated slogans for each niche, including the slogans and their corresponding niche.
- [niche-research-data]: Niche-specific research data from the Niche Researcher, including relevant keywords, phrases, and visual inspiration.
- [evaluated-slogans-list]: The slogans with their assigned humor and relevance scores in a structured format.
- [selected-slogans-list]: The selected slogans for each niche in a structured format.
- [niche]: Each niche being processed.
- [slogan]: Each slogan within a niche.
- [slogan-humor-score]: The assigned humor score (1-5) for a slogan.
- [slogan-relevance-score]: The assigned relevance score (1-5) for a slogan.
- [sorted-slogans-list]: The sorted and ranked slogans based on their combined humor and relevance scores.
- [selected-slogans-list-niche]: The top 5-10 slogans selected for a specific niche.
- [curated-slogans-list-niche]: The selected slogans compiled into a curated list for a specific niche.

subagent3refined-repeating-json-core

G) TASKFLOW SUMMARY

Title task: Evaluate slogan humor and relevance
Description of the task: GPT4 evaluates the humor and relevance of each slogan based on the niche-specific research, assigning a humor score (1-5) and a relevance score (1-5) to each slogan.
Input(s) involved: [list of generated slogans], [niche-specific research data]
Output token name: [slogans with assigned scores]

Title task: Output slogans with assigned scores
Description of the task: GPT4 outputs the slogans with their assigned humor and relevance scores in a structured format (e.g., JSON or CSV).
Input(s) involved: [slogans with assigned scores]
Output token name: [structured slogan data]

Title task: Sort and rank slogans
Description of the task: Code Interpreter sorts and ranks the slogans based on their combined humor and relevance scores.
Input(s) involved: [structured slogan data]
Output token name: [sorted and ranked slogans]

Title task: Select top slogans for each niche
Description of the task: Code Interpreter selects the top 5-10 slogans for each niche based on their combined humor and relevance scores.
Input(s) involved: [sorted and ranked slogans]
Output token name: [selected top slogans]

Title task: Output selected slogans for each niche
Description of the task: Code Interpreter outputs the selected slogans for each niche in a structured format (e.g., JSON or CSV).
Input(s) involved: [selected top slogans]
Output token name: [structured selected slogans]

Title task: Compile curated slogan list for each niche
Description of the task: GPT4 compiles the selected slogans into a curated list for each niche, ensuring the list is well-structured and easily readable.
Input(s) involved: [structured selected slogans]
Output token name: [curated slogan list]

Title task: Output final curated slogan list
Description of the task: GPT4 outputs the final curated list of the best slogans for each niche in a structured text format (e.g., JSON or CSV).
Input(s) involved: [curated slogan list]
Output token name: [final slogan list]

subagent3mermaid

prompts-required

Based on the components mentioned above, here are the unique prompts that you will need to write for this agent workflow:

Task: Niche-specific Keyword and Phrase Extraction
Output of the task: A list of relevant keywords, phrases, and visual inspiration for each niche.
Description: Write a prompt that takes the results of the Google image search for "{niche topic} birthday cards" as input and extracts relevant keywords, phrases, and visual elements that can be used for slogan generation and doodle image creation. The prompt should focus on identifying common themes, jokes, and visual styles associated with each niche.

Task: Humorous Slogan Generation
Output of the task: Multiple humorous slogans or puns for each niche.
Description: Develop a prompt that takes the niche-specific keywords and phrases as input and generates a variety of funny slogans and puns. The prompt should encourage the use of wordplay, inside jokes, and references relevant to the niche. It should also specify the desired tone (e.g., humorous, witty, or slightly irreverent) and the maximum length of each slogan (e.g., two short phrases).

Task: Best Slogan Selection
Output of the task: A curated list of the best slogans for each niche.
Description: Create a prompt that guides the AI model in selecting the most humorous, relevant, and appropriate slogans from the generated options. The prompt should provide criteria for evaluating the slogans, such as relevance to the niche, cleverness of wordplay, and overall comedic value. It should also ensure that the selected slogans are suitable for a birthday card context.

Task: Doodle Image Generation
Output of the task: Detailed image descriptions for AI-generated doodle-style images.
Description: Craft a prompt that combines the niche-specific keywords, phrases, and selected slogans to generate detailed descriptions of the desired doodle-style images. The prompt should specify the required hand-drawn aesthetic, the inclusion of relevant visual elements from the niche research, and the need for the images to complement the chosen slogans. These descriptions will be used as input for the AI image generation tool (e.g., DALL-E).

Task: Birthday Card Template Instructions
Output of the task: Clear instructions for creating birthday card templates in Bannerbear.
Description: Write a prompt that provides step-by-step instructions for designing and setting up the birthday card templates in Bannerbear. The prompt should specify the required placeholders for text and images (e.g., "card-text-line-1", "card-text-line-2", and "doodle-image-url"), as well as any additional design elements or constraints to ensure a visually appealing and consistent layout across all cards.

These prompts will help you generate the necessary content and assets for each component of the agent workflow. Remember to refine and iterate on these prompts as needed to achieve the desired results for your funny birthday card generator.

subagent1-json-array

[
  {
    "task ID": 1,
    "title of the task": "Generate list of niches",
    "description of the task": "GPT4 generates a list of 5-10 potential niches or topics related to the theme or audience based on the initial text prompt.",
    "input description": "A text prompt describing the theme or audience for birthday card ideas",
    "input required": [],
    "output description": "A list of 5-10 potential niches or topics related to the input theme or audience",
    "output name": "List of niches"
  },
  {
    "task ID": 2,
    "title of the task": "Search Google Images for niche",
    "description of the task": "Search Google Images for \"{niche topic} birthday cards\" to identify popular visual styles and gather image URLs for each niche.",
    "input description": "A single niche topic from the list generated in Task 1",
    "input required": ["List of niches"],
    "output description": "Search results containing image URLs related to the input niche topic",
    "output name": "Image search results"
  },
  {
    "task ID": 3,
    "title of the task": "Select top 5 images for niche",
    "description of the task": "GPT4 selects the top 5 most relevant and visually inspiring image URLs from the search results for each niche.",
    "input description": "Image search results from Task 2 for a single niche topic",
    "input required": ["Image search results"],
    "output description": "A list of the top 5 most relevant and visually inspiring image URLs for the niche topic",
    "output name": "Top 5 image URLs"
  },
  {
    "task ID": 4,
    "title of the task": "Save image to server",
    "description of the task": "Save each selected image URL to the local server using the \"Save external web file to server\" task.",
    "input description": "A single image URL from the top 5 selected in Task 3",
    "input required": ["Top 5 image URLs"],
    "output description": "The saved image file on the local server",
    "output name": "Saved image file"
  },
  {
    "task ID": 5,
    "title of the task": "Extract visual data from image",
    "description of the task": "Use the \"Extract visual data from image\" task to gain insights into popular elements and styles within each saved image.",
    "input description": "A single saved image file from Task 4",
    "input required": ["Saved image file"],
    "output description": "Extracted visual data containing insights into popular elements and styles within the image",
    "output name": "Extracted visual data"
  },
  {
    "task ID": 6,
    "title of the task": "Find related keywords for niche",
    "description of the task": "Use the \"Find related keywords\" task to obtain related keywords and phrases for each niche based on the initial niche description.",
    "input description": "A single niche topic from the list generated in Task 1",
    "input required": ["List of niches"],
    "output description": "A list of related keywords and phrases for the niche topic",
    "output name": "Related keywords"
  },
  {
    "task ID": 7,
    "title of the task": "Compile research for niche",
    "description of the task": "GPT4 analyzes the extracted visual data and related keywords, then compiles the research into a structured JSON format for each niche.",
    "input description": "Extracted visual data from Task 5 and related keywords from Task 6 for a single niche topic",
    "input required": ["Extracted visual data", "Related keywords"],
    "output description": "A structured JSON object containing the compiled research data for the niche topic",
    "output name": "Niche research JSON"
  },
  {
    "task ID": 8,
    "title of the task": "Combine JSON data for all niches",
    "description of the task": "GPT4 combines the JSON data for all niches into a single structured JSON file.",
    "input description": "Niche research JSON objects from Task 7 for all niche topics",
    "input required": ["Niche research JSON"],
    "output description": "A single structured JSON file containing the combined research data for all niche topics",
    "output name": "Combined research JSON"
  },
  {
    "task ID": 9,
    "title of the task": "Save final JSON file to server",
    "description of the task": "Use the \"Save content to server\" task to store the final combined JSON file on the server with a unique URL.",
    "input description": "The combined research JSON file from Task 8",
    "input required": ["Combined research JSON"],
    "output description": "A unique URL pointing to the final JSON file stored on the server",
    "output name": "Final JSON file URL"
  }
]

subagent1-json-array-mermaid

subagent2-json-array

[
  {
    "task ID": 1,
    "title of the task": "Generate custom LLM prompts",
    "description of the task": "Use the Prompt Generator (Task 2) to create 3-5 custom LLM prompts tailored to the niche based on the niche-specific research data.",
    "input description": "A JSON file containing niche-specific research data, including relevant keywords, phrases, and topics.",
    "input required": ["[niche-research-data]"],
    "output description": "A list of 3-5 custom LLM prompts specifically designed to generate humorous slogans and puns for the given niche.",
    "output name": "[custom-llm-prompts]"
  },
  {
    "task ID": 2,
    "title of the task": "Generate slogan ideas using custom prompts",
    "description of the task": "Use GPT4 (Task 1) to generate 20-30 slogan ideas for each custom prompt, employing techniques like play-on-words and puns.",
    "input description": "The custom LLM prompts generated in the previous task, which are tailored to the specific niche.",
    "input required": ["[custom-llm-prompts]"],
    "output description": "A comprehensive list of 60-150 slogan ideas generated using the custom LLM prompts, featuring humorous wordplay and puns relevant to the niche.",
    "output name": "[generated-slogan-ideas]"
  },
  {
    "task ID": 3,
    "title of the task": "Filter and refine generated slogans",
    "description of the task": "Use GPT4 (Task 1) to filter and refine the generated slogans, selecting the top 10-20 slogans per niche, each with a maximum length of 100 characters.",
    "input description": "The list of generated slogan ideas from the previous task, which includes a wide range of humorous slogans and puns.",
    "input required": ["[generated-slogan-ideas]"],
    "output description": "A curated list of the top 10-20 slogans per niche, each slogan being no longer than 100 characters and selected based on its humor, relevance, and overall quality.",
    "output name": "[refined-slogans]"
  },
  {
    "task ID": 4,
    "title of the task": "Compile refined slogans into JSON format",
    "description of the task": "Use GPT4 (Task 1) to compile the refined slogans into a structured JSON format, organizing the slogans by niche.",
    "input description": "The curated list of top 10-20 slogans per niche, which have been filtered and refined based on humor, relevance, and quality.",
    "input required": ["[refined-slogans]"],
    "output description": "A well-structured JSON file containing the top 10-20 humorous slogans for each niche, organized in a clear and easily accessible format.",
    "output name": "[slogan-json-output]"
  }
]

2. Make funny birthday cards for niche

Sub Agent #1 | Sub Agent #2 | Sub Agent #3

final-output-comps..1

final-output-comps..2

final-output-comps GPT..3

CLAUDE:- final-output-and-component-outputs

claude 3 - extract final output

extract component outputs

Claude - research-ideas-and-input-variables

GPT - research-ideas-and-input-variables

template-ideas

claude 3 - searches

subagent-summary

subagent-summary-amended-for-actual-tasks..1

subagent-summary-amended-for-actual-tasks..2

subagent-summary-amended-for-actual-tasks GPT..3

subagent-summary-amended-for-actual-tasks

random-autocomplete

claude 3 - mermaid-for-all

SubAgent #1

claude 3 - subagent1

subagent1-refined

subagent1refined-repeating

Subagent1 task-ID-numbers

subagent1-output-variables

subagent1refined-repeating-json-core

subagent1mermaid

SubAgent #2

claude 3 - subagent2

subagent2-refined

subagent2refined-repeating

Subagent2 task-ID-numbers

subagent2-output-variables

subagent2refined-repeating-json-core

subagent2mermaid

SubAgent #3

claude 3 - subagent3

subagent1-refined

subagent1refined-repeating

Subagent1 task-ID-numbers3

subagent3-output-variables

subagent3refined-repeating-json-core

subagent3mermaid

prompts-required

subagent1-json-array

subagent1-json-array-mermaid

subagent2-json-array

subagent2-json-array-mermaid