No Clocks Vault

Guide - AI Chat Processing Workflow

5 min read

Guide - AI Chat Processing Workflow

Overview

This guide walks through processing AI LLM chat exports and extracting valuable knowledge into atomic notes for your Zettelkasten. It covers the complete workflow from capture to archival across multiple AI providers.

Prerequisites

  • Obsidian vault with Templater plugin configured
  • Templates: Template-AIChat.md, Template-Insight.md
  • Optional: Nexus AI Chat Importer plugin (for ChatGPT/Claude ZIP exports)
  • Optional: Atomizer plugin (for AI-powered extraction)

The CEPA Framework

The AI Chat processing workflow follows four phases:

PhaseActionOutput
CaptureImport/paste raw chat exportRaw chat note in Inbox
EvaluateReview for valuable insightsPrioritized extraction list
ProcessExtract atomic notesDefinition/Code/Insight notes
ArchiveLink and file completed chatOrganized knowledge graph

Phase 1: Capture

Manual Export (Perplexity, Claude, General)

  1. Export conversation from the AI provider (usually as Markdown)
  2. Create new note using Template-AIChat.md
  3. Name using convention: YYYY-MM-DD_Provider_Topic.md
  4. Place in 04-RESOURCES/Integrations/Chats/
  5. Paste raw content into the “Chat Content” section
  6. Update frontmatter: provider, chat_date, topics

Automated Import (ChatGPT)

For bulk ChatGPT imports, use the Nexus AI Chat Importer plugin:

  1. Export data from ChatGPT (Settings → Data Controls → Export)
  2. Open Command Palette → “Nexus AI: Import conversations”
  3. Select the downloaded ZIP file
  4. Configure target folder: 04-RESOURCES/Integrations/Chats/
  5. Enable date prefix for organization

Perplexity-Specific Format

Perplexity exports include citations in footnote format. Preserve these:

Some claim about information theory.[^1_1]
 
[^1_1]: https://example.com/source

The citations provide valuable sources for verification and further reading.

Phase 2: Evaluate

Triage Questions

For each chat, ask:

  1. Novelty: Does this contain ideas I haven’t captured before?
  2. Quality: Is the information accurate and well-reasoned?
  3. Actionability: Can I apply this knowledge or connect it to existing notes?
  4. Density: How many distinct insights are worth extracting?

Value Assessment Matrix

ScoreCriteriaAction
HighMultiple novel insights, high confidenceFull extraction
MediumSome useful content, mixed qualitySelective extraction
LowMostly review of known infoQuick scan, minimal extraction
NoneCasual chat, troubleshooting onlyArchive as-is

Marking Insights

As you read, update the “Key Insights” section:

## Key Insights
 
> [!TIP] Insights to Extract
> 
> - [x] Shannon's information theory as physics bridge → [[Definition - Information Theory]]
> - [x] Wheeler's "it from bit" concept → [[Wheeler It From Bit]]
> - [ ] Bekenstein-Hawking entropy formula → [[Definition - Bekenstein Bound]]
> - [ ] Landauer's principle → [[Definition - Landauer Principle]]

Phase 3: Process

Extraction Strategy

Match content type to target note type:

Content TypeTarget TemplateTarget Folder
Term/Concept definitionTemplate-Definition.md04-RESOURCES/Definitions/
Code snippet/exampleTemplate-Code.md04-RESOURCES/Code/{Lang}/
Novel insight/ideaTemplate-Insight.md01-SLIPBOX/
How-to procedureTemplate-Guide.md04-RESOURCES/Guides/
Tool/resource listTemplate-List.md04-RESOURCES/Lists/

Creating Atomic Notes

For each insight:

  1. Create new note with appropriate template
  2. Use clear, searchable title (no chat reference in title)
  3. Write the core concept in your own words
  4. Add source_chat link in frontmatter
  5. Tag with relevant Topic/ tags
  6. Link to related existing notes

Example: Extracting a Definition

From the Perplexity chat about information theory:

Source text:

“Wheeler’s ‘it from bit’ directly addresses your observation: every item of the physical world has at bottom—at a very deep bottom, in most instances—an immaterial source and explanation.”

Extracted note: Definition - It From Bit.md

---
creation_date: 2024-12-31
tags:
  - Type/Definition
  - Topic/Physics
  - Topic/InformationTheory
source_chat: "[[2025-12-31_Perplexity_Information theory computational encoding and quantum physics]]"
---
 
# Definition - It From Bit
 
## Overview
 
**It from Bit** is a phrase coined by physicist John Archibald Wheeler proposing that information is fundamental to physics—that physical entities derive their existence from binary choices and information rather than the reverse.
 
## Key Concepts
 
- Physical reality emerges from informational acts (observation, measurement)
- Particles, fields, and spacetime are manifestations of underlying information
- Binary yes/no questions form the substrate of physical existence
 
## See Also
 
- [[Definition - Information Theory]]
- [[Definition - Bekenstein Bound]]

Batch Processing with AI

For dense chats, consider the Atomizer plugin:

  1. Open Command Palette → “Atomize Selected Note”
  2. Configure OpenAI API key in settings
  3. Review generated atomic notes
  4. Merge, edit, or discard as appropriate
  5. Add proper frontmatter and links

Phase 4: Archive

Update Source Chat

After extraction, update the original chat note:

  1. Change extraction_status to complete
  2. Add extracted note links to frontmatter array:
extracted_notes:
  - "[[Definition - It From Bit]]"
  - "[[Definition - Bekenstein Bound]]"
  - "[[Landauer Principle and Computation]]"
  1. Check off items in “Key Insights” section
  2. Add completion date to “Processing Notes”

Filing Decisions

Chat ValueDestinationRationale
High-referenceKeep in 00-INBOX/Chats/May need to revisit
Fully extractedMove to 99-ARCHIVES/AI-Chats/Knowledge captured elsewhere
Low valueDelete or 99-ARCHIVES/Minimal ongoing value

Creating Archive Folder

If it doesn’t exist:

99-ARCHIVES/
└── AI-Chats/
    ├── 2024/
    │   ├── 12/
    │   └── ...
    └── 2025/
        └── ...

Automation Ideas

Dataview Dashboard

Add to your daily note or PKM dashboard:

QuickAdd Macro

Create a QuickAdd macro for rapid insight capture:

  1. Prompt for insight title
  2. Prompt for source chat (suggester from Chats folder)
  3. Create note from Template-Insight.md
  4. Open in edit mode

Weekly Review Checklist

  • Process any new chat exports in Inbox
  • Review extraction_status: in_progress chats
  • Archive fully-processed chats
  • Run Smart Connections to find new links

Provider-Specific Tips

Perplexity

  • Preserves citations with footnote syntax
  • Often includes multiple “turns” separated by ---
  • Sources are valuable—consider extracting to a References list

ChatGPT

  • Use Nexus AI Chat Importer for bulk processing
  • Conversation titles may need renaming
  • Watch for markdown formatting inconsistencies

Claude

  • Clean markdown export
  • Artifacts (code blocks) are well-preserved
  • Consider extracting artifact code separately

Cursor/Copilot

  • Context is often code-specific
  • Extract working solutions to Code notes
  • Document the problem-solution pattern

Troubleshooting

Markdown Formatting Issues

Problem: Code blocks or LaTeX not rendering correctly.

Solution:

  • Ensure proper fence syntax (triple backticks with language)
  • For LaTeX, use $...$ for inline, $$...$$ for block
  • Some exports use \(...\) which may need conversion

Duplicate Imports

Problem: Same conversation imported multiple times.

Solution:

  • Nexus AI tracks processed ZIPs; don’t re-process
  • Check for existing notes before creating duplicates
  • Use file dates in names to distinguish versions

Missing Context

Problem: Extracted note loses important context from conversation flow.

Solution:

  • Include more surrounding context in extraction
  • Add explicit source_chat link for reference
  • Quote key passages directly in the atomic note

Appendix

Created: 2024-12-31 | Modified: 2024-12-31

See Also

(c) No Clocks, LLC | 2024

Graph View

Backlinks