ChatGPT Memory Management Workflow

Error converting content: marked is not a function

type:: meta-system

zettel-id:: 25h
parent-id:: 25c
created:: 2025-05-12
- # ChatGPT Memory Management Workflow
- ## Purpose
  A repeatable protocol for turning ChatGPT sessions into structured Logseq knowledge—so insights never vanish in scrollback.
- ## Pipeline
  1. **Session Tagging**
- Title every chat export as `# YYYY-MM-DD – Session Summary: `.
- Assign `zettel-id` and link to `parent-id` (previous related note).
	  
	  2. **Session Note Extractor Prompt**
- Run extractor to generate: Highlights, Permanent Notes, Open Questions, Link Suggestions.
- Output Markdown only → paste into Logseq page.
- Prompt:
	  ```
	  You are my "Session Note Extractor." Take the entire conversation that just finished and output a single Markdown file, formatted for Logseq, that follows this structure:
	  
	  1. Title line = # YYYY-MM-DD – Session Summary: 
	  
	  2. Page properties table:
	     type:: session
	     tags:: chatgpt-memory areas/
	     zettel-id:: <% check ID Ledger for next ID in sequence %>
	     parent-id:: <% most recent related zettel-id, if any %>
	     created:: YYYY-MM-DD
	  
	  3. ## Highlights (Fleeting Notes)
	     * Bullet list of 5–15 key takeaways ONLY those topics that will serve as long term memory for AI.
	     * When applicable, Use EXACT phrasing from our conversation (just trim filler words)
	     * Do NOT add new ideas or knowledge not covered in our chat
	     
	  4. ## Permanent Notes
	     * Transform highlights into concise notes STRICTLY based on what we discussed
	     * One idea per bullet, linking ONLY to relevant Logseq pages mentioned or directly related
	     * Add "why this matters" commentary ONLY if we discussed the importance
	     * NEVER add speculative content or ideas not explicitly covered
	     
	  5. ## Open Questions / TODO
	     * ONLY include action items or questions that explicitly surfaced in our conversation
	     * Format as checkboxes
	     
	  6. ## Link Suggestions
	     * ONLY include pages that were directly mentioned or highly relevant to our specific discussion
	     * Do NOT suggest tangential or speculative connections
	  
	  Output ONLY the Markdown based STRICTLY on our actual conversation, nothing else.
	  ```
	  
	  3. **Page Creation Routine**
- One Logseq page per major link in "Link Suggestions" (appearance, skincare, etc.).
- Use page template:  
	  ```
	  # Page Title
	  type:: concept / project / routine
	  zettel-id:: XX
	  parent-id:: YY
	  created:: YYYY-MM-DD
	  ```
	  
	  4. **Incremental ID Scheme**
- Alphanumeric suffix (25a, 25b, 25c…) for quick reference in chat.
- Store latest `zettel-id` in a running index page `ID Ledger`.
	  
	  5. **Two-Week Refresh Cycle**
- Bi-weekly ChatGPT doc update: add new aliases, update statuses, correct brand swaps.
- Use "Update cadence" field at top of each page.
	  
	  6. **Link Hygiene**
- Minimum three backlinks per new page to existing concepts.
- Use `` liberally so graph view stays dense.
	  
	  7. **Automation Wishlist**
- Future: script to auto-push extractor output to Logseq via Git commit.
- Logseq Smith package for programmatic file stitching and Logseq API integration.
- Tag tasks under `Automation Ideas`.
- ## Current Pages Using This Workflow
- Appearance Systems
- Wardrobe Experiments 2025
- Skincare Systems
- Footwear Rotation
- Fabric Tactics
- ## Reference Page Creation
- After generating the session summary, create individual reference pages one at a time:
  ```
  Based on the session summary you created, I'd like you to help me create or update a reference page for SPECIFIC PAGE NAME. 
  
  Please format this as a Logseq markdown file with:
  
  1. Page title = # SPECIFIC PAGE NAME
  
  2. Properties table:
     type:: reference
     tags:: chatgpt-memory areas/ 
     zettel-id:: <% next ID in sequence from ID Ledger %>
     parent-id:: <% session zettel-id %>
     created:: 
  
  3. ## Core Concept
     * Concise 1-2 sentence definition based ONLY on what we discussed
     
  4. ## Key Points
     * Extract and organize the relevant information from our conversation
     * Include only information we explicitly covered about this topic
     * Format in clear, scannable bullet points
  
  5. ## Implementation
     * Present any practical applications we discussed as actionable steps
     * Integrate links to related systems naturally within the content
     * Focus on the "how" aspects we covered
  
  Use ONLY information from our actual discussion without adding speculation or external knowledge. Keep the page focused, practical, and directly tied to our conversation.
  ```
- Review each page before requesting the next one from the Link Suggestions section
- Maintain strict adherence to information actually discussed
- Integrate links naturally within content rather than as separate sections
- ## Next Actions
- DONE Create `ID Ledger` and log current highest zettel-id (now 25p).
- TODO Schedule bi-weekly "doc refresh" automation (May 19).
- DONE Draft extractor prompt as reusable snippet in ChatGPT custom instructions.
- TODO Update total focus time metrics (8.5h on 2025-05-12).
- ## Related Links
- Knowledge Management Workflows
- Automation Ideas