Ebook from Title and Keywords

I want to create a short ebook on any topic. The idea is I enter a title, vague sections of what I want covered, perhaps 3-5 keywords to research, then the AI will go ahead and research these keywords using a Google search, select (say) 3 URLs for each keyword, crawl them, summarize them. Then the AI will review the researched content, decide on say 10 specific sections, and then write each section one by one. Lastly, the AI will use a text to PDF tool to turn that text into a PDF.

First, pre-entered info...

final-output2

The final output of this automated workflow would be a PDF file containing a short ebook on a specific topic chosen by the user. The ebook will have a title page, table of contents, and approximately 10 well-structured sections, each covering a specific aspect of the topic based on the AI's research using the provided keywords and selected URLs. The PDF will be formatted consistently throughout, ensuring readability and visual appeal.

subagents-extracted2

OVERVIEW OF SUBAGENTS: 1. Research Agent: Generates search queries, performs Google searches, selects top URLs, crawls content, and summarizes it 2. Content Structuring Agent: Analyzes summarized content, decides on 10 specific sections, and creates a structured outline 3. Content Generation Agent: Generates well-structured and coherent written content for each section and selects relevant images 4. PDF Generation Agent: Formats the written content, incorporates images, and converts it into a PDF document SUBAGENT 1: Research Agent - Name: ResearchConductor - Final Output: JSON object containing the search queries (list of strings), selected URLs (list of strings), crawled content (list of strings), and summaries (list of strings) - SUBAGENT FINAL OUTPUT: [research-output] - Skills Used: - #181 - Research Topic (Or Keyword) Deeply - DESCRIPTION: This skill will research the user-provided keywords and vague sections deeply, running searches on various search engines, visiting websites, and collating research information. It will then use a LLM to summarize the research based on the information requested. - INPUT NEEDED: keywords and/or topics to research (from user input) - OUTPUT: 1000 characters of text per keyword/topic - #185 - Brainstorm Most Relevant Domains/Websites - DESCRIPTION: This skill inputs a single keyword (or topic, product description, or any piece of text really) and will then run multiple searches (using keyword tools like SEMRush along with Google related suggestions, etc), and then return a list of the 4 most related domains/website URLs in order of relevancy, along with additional commentary on each (as written by a LLM). - INPUT NEEDED: single keyword (or topic, product description, or anything) - OUTPUT: list of 4 related domains, listed in order, with comments - Specific Sequence: 1. Use #181 to research the user-provided keywords and vague sections deeply 2. Use #185 to brainstorm the most relevant domains/websites based on the research 3. Select the top 3 URLs for each keyword from the brainstormed list 4. Crawl the content from the selected URLs using BeautifulSoup (REQUIRED) 5. Summarize the crawled content using a text summarization API or library (REQUIRED) 6. Store the search queries, selected URLs, crawled content, and summaries in a JSON object 7. Pass the JSON object to the next subagent SUBAGENT 2: Content Structuring Agent - Name: ContentStructurer - Final Output: JSON object containing the 10 specific section titles (list of strings) and a structured outline (dictionary with section titles as keys and bullet points or short descriptions as values) - SUBAGENT FINAL OUTPUT: [structured-outline] - Skills Used: - #238 - Code interpreter (REQUIRED) - DESCRIPTION: This skill will analyze the summaries from the previous research agent and run clustering or topic modeling algorithms to decide on 10 specific sections for the ebook, creating a structured outline. - INPUT NEEDED: Text summaries of researched content - OUTPUT: JSON object with identified sections and corresponding summarized bullet points or short descriptions - Specific Sequence: 1. Analyze the summarized research content using NLP techniques like topic modeling or clustering with #238 2. Decide on the 10 specific section titles based on the analysis 3. Create a structured outline with section titles as keys and bullet points or short descriptions as values 4. Store the section titles and structured outline in a JSON object 5. Pass the JSON object to the next subagent SUBAGENT 3: Content Generation Agent - Name: ContentCreator - Final Output: Plain text files containing well-structured and coherent written content for each of the 10 sections, along with relevant images - SUBAGENT FINAL OUTPUT: [generated-content] - Skills Used: - #177 - Write Text (Or Copy) From Researched Text - DESCRIPTION: This skill takes the structured outline and summarized research content for each section and uses a LLM (typically GPT4) to generate well-structured and coherent written content. - INPUT NEEDED: Structured outline, summarized research content for each section - OUTPUT: Plain text file with written content for each section - #169 - Create Dalle Image (Rectangle) - DESCRIPTION: This skill creates a 1792x1024 square image with DalleE, based on a given text prompt. It will be used to generate images relevant to each section to insert into the ebook for visual enhancement. - INPUT NEEDED: Text prompts related to each section for image generation (e.g., "The great barrier reef underwater scene") - OUTPUT: Internal and external link to the image - Specific Sequence: 1. For each section in the structured outline: a. Retrieve relevant information from the summarized research content b. Use #177 to generate well-structured and coherent written content c. Use #169 to generate a relevant image based on the section content 2. Store the generated content for each section in plain text files 3. Store the generated images with their respective section titles 4. Pass the generated content and images to the next subagent SUBAGENT 4: PDF Generation Agent - Name: PDFGenerator - Final Output: A well-formatted PDF file containing the ebook with a title page, table of contents, the generated content for each section, and relevant images - SUBAGENT FINAL OUTPUT: [ebook-pdf] - Skills Used: - #227 - Format Up To 4000 Characters Of Text Into Pretty HTML - DESCRIPTION: This skill formats the written content using HTML, including titles, paragraphs, bolding, bullet points, etc. It also incorporates the generated images into the respective sections. - INPUT NEEDED: Written content for each section, formatted using HTML, and the generated images - OUTPUT: HTML content ready for PDF conversion - #203 - Convert HTML URL To PDF - DESCRIPTION: This skill converts the formatted HTML content into a PDF document. - INPUT NEEDED: URL of the HTML content (the output from the previous step) - OUTPUT: PDF file of the ebook - Specific Sequence: 1. Use #227 to format the written content for each section into pretty HTML, including titles, paragraphs, bolding, bullet points, etc., and incorporate the relevant images 2. Create a title page and table of contents in HTML format 3. Concatenate the formatted content, title page, and table of contents into a single HTML file (REQUIRED) 4. Use #203 to convert the HTML file into a PDF document 5. Apply any necessary styling to ensure a visually appealing and professional-looking ebook (REQUIRED) 6. Save the final PDF file

subagentnumber

[subagentnumber]

Then, agent response...

subagent1refined-new-json-core

G) TASKFLOW SUMMARY - ID: #181 - Title task: Research Topic (Or Keyword) Deeply - Description of the task: This skill will research the user-provided keywords and vague sections deeply, running searches on various search engines, visiting websites, and collating research information. It will then use a LLM to summarize the research based on the information requested. - Input(s) involved: keywords and/or topics to research (from user input) - Output token name: 1000 characters of text per keyword/topic - ID: #185 - Title task: Brainstorm Most Relevant Domains/Websites - Description of the task: This skill inputs a single keyword (or topic, product description, or any piece of text really) and will then run multiple searches (using keyword tools like SEMRush along with Google related suggestions, etc), and then return a list of the 4 most related domains/website URLs in order of relevancy, along with additional commentary on each (as written by a LLM). - Input(s) involved: single keyword (or topic, product description, or anything) - Output token name: list of 4 related domains, listed in order, with comments - ID: (REQUIRED) - Title task: Crawl Content from Selected URLs using BeautifulSoup - Description of the task: This task uses the BeautifulSoup library to crawl and extract content from the top 3 selected URLs for each user-provided keyword and section. - Input(s) involved: selected URLs (list of strings) - Output token name: crawled content (list of strings) - ID: (REQUIRED) - Title task: Summarize Crawled Content using Text Summarization API or Library - Description of the task: This task uses a text summarization API or library to summarize the crawled content from the selected URLs for each user-provided keyword and section. - Input(s) involved: crawled content (list of strings) - Output token name: summaries (list of strings) - ID: #238 - Title task: Code Interpreter - Description of the task: This skill will analyze the summaries from the previous research agent and run clustering or topic modeling algorithms to decide on 10 specific sections for the ebook, creating a structured outline. - Input(s) involved: Text summaries of researched content - Output token name: JSON object with identified sections and corresponding summarized bullet points or short descriptions

subagent1mermaid

<!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[User Input: Keywords and Vague Sections] --> B{ResearchConductor} B -->|Keyword Research| C(Research keywords deeply using skill #181) B -->|Section Research| D(Research sections deeply using skill #181) C --> E(Brainstorm relevant domains for keywords using skill #185) D --> F(Brainstorm relevant domains for sections using skill #185) E --> G(Select top 3 URLs for each keyword) F --> H(Select top 3 URLs for each section) G --> I{Content Crawling and Summarization} H --> I I -->|Crawl content| J(Crawl content from selected URLs using BeautifulSoup) J --> K(Summarize crawled content using text summarization API or library) K --> L{JSON Object Creation} L --> M(Store search queries, selected URLs, crawled content, and summaries in a JSON object using skill #238) M --> N[Final Output: JSON object] </div> <script> mermaid.initialize({ startOnLoad: true }); </script> </body> </html>

Mermaid Image - https://static.aiz.ac/1721810444-mermaid/mermaid1.png

prompts-required

[prompts-required]

subagent1-json-array

[ { "taskID": 181, "semanticTitleOfTask": "Research User-Provided Keywords and Sections", "taskDescription": "This skill researches the user-provided keywords and vague sections deeply, running searches on various search engines, visiting websites, and collating research information. It then uses a LLM to summarize the research based on the information requested.", "inputDescription": "User-provided keywords (list of strings) and user-provided vague sections (list of strings)", "inputRequired": [], "outputDescription": "1000 characters of text per keyword/topic, summarizing the deep research conducted", "outputName": "keyword-section-research", "promptInstruction": "Your task is to research the following keyword(s) or topic(s) in depth: `keywords_or_topics`. Conduct searches on various search engines, visit relevant websites, and collate the most important information. Summarize your findings in approximately 1000 characters per keyword/topic, focusing on the key points and insights. Your summary should provide a comprehensive overview of each keyword/topic based on your extensive research." }, { "taskID": 185, "semanticTitleOfTask": "Brainstorm Relevant Domains for Each Keyword/Section", "taskDescription": "This skill inputs a single keyword (or topic, product description, or any piece of text) and runs multiple searches (using keyword tools like SEMRush along with Google related suggestions, etc), returning a list of the 4 most related domains/website URLs in order of relevancy, along with additional commentary on each (as written by a LLM).", "inputDescription": "Research summaries for each user-provided keyword and section from the previous task", "inputRequired": ["keyword-section-research"], "outputDescription": "List of 4 related domains for each keyword/section, listed in order, with comments", "outputName": "relevant-domains", "promptInstruction": "Given the following keyword, topic, or text: `input_text`, your task is to identify the 4 most relevant domains or websites. Use keyword research tools like SEMRush and Google's related search suggestions to find the most appropriate websites. List the 4 domains in order of relevance, and provide a brief commentary on why each domain is relevant to the given input. Your output should help users quickly identify the best resources for further information on the given keyword, topic, or text." }, { "taskID": 1, "semanticTitleOfTask": "Select Top 3 URLs for Each Keyword/Section", "taskDescription": "This task selects the top 3 URLs from the list of relevant domains for each user-provided keyword and section.", "inputDescription": "List of 4 related domains for each keyword/section from the previous task", "inputRequired": ["relevant-domains"], "outputDescription": "Top 3 selected URLs for each user-provided keyword and section", "outputName": "selected-urls", "promptInstruction": "" }, { "taskID": 2, "semanticTitleOfTask": "Crawl Content from Selected URLs using BeautifulSoup", "taskDescription": "This task uses the BeautifulSoup library to crawl and extract content from the top 3 selected URLs for each user-provided keyword and section.", "inputDescription": "Top 3 selected URLs for each user-provided keyword and section from the previous task", "inputRequired": ["selected-urls"], "outputDescription": "Crawled content from the selected URLs for each user-provided keyword and section", "outputName": "crawled-content", "promptInstruction": "" }, { "taskID": 3, "semanticTitleOfTask": "Summarize Crawled Content using Text Summarization API or Library", "taskDescription": "This task uses a text summarization API or library to summarize the crawled content from the selected URLs for each user-provided keyword and section.", "inputDescription": "Crawled content from the selected URLs for each user-provided keyword and section from the previous task", "inputRequired": ["crawled-content"], "outputDescription": "Summaries of the crawled content for each user-provided keyword and section", "outputName": "content-summaries", "promptInstruction": "Your task is to summarize the content crawled from the following URLs: `crawled_content`. Use advanced text summarization techniques to extract the most important information and condense it into concise summaries. Each summary should capture the key points and main ideas from the corresponding URL's content. Aim to provide a clear and informative overview of the content in a compact format. Your summaries will be used to create a structured outline for an ebook on the given topic." }, { "taskID": 238, "semanticTitleOfTask": "Create Structured JSON Object with Identified Sections and Summaries", "taskDescription": "This skill analyzes the summaries from the previous research tasks and runs clustering or topic modeling algorithms to decide on 10 specific sections for the ebook, creating a structured outline.", "inputDescription": "Summaries of the crawled content for each user-provided keyword and section from the previous task", "inputRequired": ["content-summaries"], "outputDescription": "JSON object with identified sections and corresponding summarized bullet points or short descriptions", "outputName": "structured-outline", "promptInstruction": "Analyze the following text summaries: `text_summaries`, which contain researched content on a specific topic. Your task is to identify the 10 most important sections or subtopics that should be included in an ebook about this topic. Use natural language processing techniques like topic modeling or clustering to group similar ideas and identify the key themes. Create a structured JSON object with the identified section titles as keys and corresponding summarized bullet points or short descriptions as values. Your output will serve as the foundation for generating a well-organized ebook on the given topic." } ]

subagent1-json-array-mermaid

<!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[User Input: Keywords and Sections] --> B{Research User-Provided Keywords and Sections} B --> C{Brainstorm Relevant Domains for Each Keyword/Section} C --> D{Select Top 3 URLs for Each Keyword/Section} D --> E{Crawl Content from Selected URLs using BeautifulSoup} E --> F{Summarize Crawled Content using Text Summarization API or Library} F --> G{Create Structured JSON Object with Identified Sections and Summaries} G --> H[Final Output: JSON Object with Research Data] </div> <script> mermaid.initialize({ startOnLoad: true }); </script> </body> </html>