claude-code-ppt-generation-team CLAUDE.md

Interaction Relationships and Call Flow (Pipeline)

1. Start:

   • User ==> presentation-orchestrator: Submit the folder path containing all photos.

2. Phase One: Image Analysis

   • presentation-orchestrator ==> image-analyzer: "Please process each photo in this folder and generate corresponding Markdown files and reconstructed charts for each photo."
   • image-analyzer ==> presentation-orchestrator: Return a list of all .md files and reconstructed chart image paths.

3. Phase Two: Content Synthesis & Storytelling

   • presentation-orchestrator ==> content-synthesizer: "Please read these Markdown files, organize them into a logical presentation outline, and add your insights."
   • content-synthesizer ==> presentation-orchestrator: Return an outline file that defines the presentation structure.

4. Phase Three: Presentation Generation

   • presentation-orchestrator ==> presentation-designer: "Please create a PowerPoint presentation based on this outline and all related materials (photos, charts)."
   • presentation-designer ==> presentation-orchestrator: Return a draft .pptx file.

5. Phase Four: Quality Assurance & Delivery

   • presentation-orchestrator ==> quality-assurance-editor: "Please review this presentation to ensure content and format quality."
   • quality-assurance-editor ==> presentation-orchestrator: Return the final .pptx file.

6. End:

   • presentation-orchestrator ==> User: Deliver the final presentation file.

Workspace Environment Data Architecture Blueprint

This is the standard structure that presentation-orchestrator should initialize when receiving a new presentation creation project (e.g., project_exhibition_summary).

📁 project_exhibition_summary/
│
├── 📄 README.md                 # Project documentation maintained by Orchestrator
│
├── 📁 00_inputs/                 # Original data provided by user
│   └── 📁 raw_photos/           # Store all exhibition photos (immutable)
│       ├── 📄 IMG_001.jpg
│       └── 📄 IMG_002.png
│
├── 📁 01_processing_outputs/    # All intermediate processing outputs
│   ├── 📁 extracted_markdown/   # MD files generated by image-analyzer
│   │   ├── 📄 IMG_001.md
│   │   └── 📄 IMG_002.md
│   └── 📁 reconstructed_figures/ # Tables and flowcharts reconstructed by image-analyzer
│       └── 📄 IMG_002_table.png
│
├── 📁 02_src/                     # Store reusable Python source code
│   ├── 📄 __init__.py
│   ├── 📄 content_synthesis.py  # Core functions for content-synthesizer (text analysis, structure planning)
│   └── 📄 pptx_generator.py     # Core functions for presentation-designer (automated PPT generation)
│
└── 📁 03_deliverables/            # Final deliverables for user
    ├── 📄 presentation_outline.md # (Optional) Final outline generated by content-synthesizer for reference
    └── 📄 exhibition_summary_final.pptx # Final presentation file output

Python Execution Environment

• C:/Users/hunghsun/AppData/Local/anaconda3/envs/py10/python.exe
• Key Library Recommendations:
  • Image Processing & OCR: Use Claude Sonnet's own model for recognition
  • Chart Reconstruction: matplotlib, pandas, seaborn, graphviz
  • Presentation Generation: python-pptx

Important Notes

• 1. If encountering photos that are difficult to recognize, ignore them and do not perform analysis or generate MD files, nor include them in the PPT file
• 2. Interpret from a technical perspective, avoid personal emotional expression. Photos are usually from AI-related technical seminars or vendor AI product demonstrations
• 3. Write in a tech article style for introducing meeting content and new knowledge to supervisors
• 4. There will be an existing folder responsible for storing all user photos, with relative path: .\00_inputs\raw_inputs
• 5. If encountering photos with extremely similar content, do not write repetitive article content

Therefore, there is no need to create this folder additionally, just create the other folders.