Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Video2Post Generation Workflow #49

Draft
wants to merge 5 commits into
base: master
Choose a base branch
from
Draft

Conversation

xihajun
Copy link

@xihajun xihajun commented Feb 26, 2024

(will add to minbpe-doc)

graph TD;
    classDef success fill:#5CB85C,stroke:#fff,color:#fff;
    classDef progress fill:#428BCA,stroke:#fff,color:#fff;
    classDef pending fill:#F0AD4E,stroke:#fff,color:#fff;
    
    VideoScript[Video Scripts] --> OutlineGeneration[Generate Outline];

    subgraph IntegrationAndPreparation_Claude
        VideoScript --> PromptConstruction[Construct Prompt];
        OutlineGeneration --> PromptConstruction;
        CodeSnippet[Code Snippets] --> PromptConstruction;
        PromptConstruction --> FewShotExamples[Incorporate Few-Shot Examples];
    end
    
    subgraph FirstSection
        FewShotExamples --> GenSection1[Section 1: Generated Content];
    end
    
    subgraph SubsequentSections
        GenSection1 --> |Next Section Input| GenSection2[Section X: Generated Content];
    end
    
    subgraph IterativeReview_GPT4
        Reviewer -->|Provides Feedback| Writer;
        Writer -->|Implements Modifications| Reviewer;
    end
    
    SubsequentSections --> IterativeReview_GPT4;
    IterativeReview_GPT4 --> SubsequentSections;
    FirstSection --> IterativeReview_GPT4;
    IterativeReview_GPT4 --> FirstSection;
    GenSection2 --> FinalOutput[Combine All Sections];
    
    class GenSection1,GenSection2 progress;
    class GenSection2 pending;
    class FinalOutput success;
    style IterativeReview_GPT4 color:#f66
Loading

Optimisation Principle

  • Simplify Agent Collaboration:
    • Minimize collaboration layers.
    • Enhance data sharing efficiency. (Data is easy to move, while an agent is hard to design.)
      • Cooking a better tool for agent communication (over engineering)

New Features

  • Fact-Checker Is All You Need (eval):
    • ✅ 'Reviewer Mode' for quality checks.
    • ✅ Integrate Screenshot Support (@sebasibarguen's script, OCR, OpenCV, MP4)
    • How to evaluate the generated results? Human reviewer? functional call searching?
  • Tested on Claude-3
  • Add human-in-loop Feature

Good Tutorial

  • Clear Goals for Each Page (focus)
  • Visual Aids (iiagrams, flowcharts)
  • Iteractive Element
  • Active Learning (ask why, read more)
    • AI-Powered Discussion
    • AI reading list
    • Project/Application Oriented (problem solving)
  • Gamified
    • Error Discovery (Intentional Errors Identification Game)
    • Rewards
    • Leaderboards
  • 10000 hours countdown

@karpathy
Copy link
Owner

Did you write this out yourself? :)

@xihajun
Copy link
Author

xihajun commented Feb 27, 2024

Haha, it depends! Does it pass the Turing test, or the Karpathy test :)

An updated version is coming soon. Would love to hear your feedback!

@xihajun
Copy link
Author

xihajun commented Mar 4, 2024

"Fully automated example" using GitHub action and designed AI Agents:

image

I think I am making some progress on the automation pipeline; next up is establishing a clear definition of good content :)

Fully automated?
image

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants