LLM Logs to Obsidian

A complete system for integrating Claude Code with Obsidian - conversation logging, imports, and organized knowledge management.

What's Included

Component	Description
Import Scripts	Import ChatGPT history and export Claude Code transcripts
CLAUDE.md Template	Instructions for Claude to log sessions automatically
Vault Structure	Ready-to-use Obsidian folder structure
Documentation	Setup guides and troubleshooting

Quick Start

1. Clone this repo:

   git clone https://github.com/DiamondEyesFox/llm-obsidian-scripts.git
   cd llm-obsidian-scripts

2. Copy vault structure to your Obsidian vault:

   cp -r templates/vault-structure/Claude\ Dashboard/ ~/Documents/YourVault/

3. Set up CLAUDE.md:

   cp templates/CLAUDE.md ~/CLAUDE.md
   # Edit and replace {{VAULT_PATH}} with your vault path

4. Configure scripts:

   mkdir -p ~/.claude/hooks
   cp scripts/raw_log.js ~/.claude/hooks/
   # Edit and set VAULT_ROOT in the script

See docs/SETUP.md for detailed instructions.

The Dual Logging System

This system uses two complementary logs:

Log Type	Purpose	Created By
Summary Logs	Quick reference, searchable	Claude (auto, per CLAUDE.md)
Raw Transcripts	Full history, debugging	raw_log.js (manual export)

Important: Neither fully auto-fires. Summary logs depend on Claude following CLAUDE.md instructions. Raw exports require running a command after sessions.

See docs/SESSION_LOGGING.md for details.

Repository Structure

llm-obsidian-scripts/
├── scripts/
│   ├── chatgpt-import.py    # Import ChatGPT exports
│   └── raw_log.js           # Export Claude Code transcripts
├── templates/
│   ├── CLAUDE.md            # Claude Code instructions
│   └── vault-structure/     # Obsidian folder template
│       └── Claude Dashboard/
├── docs/
│   ├── SETUP.md             # Installation guide
│   ├── SESSION_LOGGING.md   # How logging works
│   └── TROUBLESHOOTING.md   # Common issues
└── examples/
    └── sample-session-log.md

Vault Structure

After setup, your vault will have:

Your Vault/
├── Claude Dashboard/
│   ├── Inbox/           # Files for Claude to process
│   ├── Outbox/          # Files Claude creates
│   ├── Reference/       # Persistent context
│   ├── Session Logs/    # Daily summaries
│   ├── Scratchpad.md    # Working memory
│   └── Todo.md          # Pending tasks
└── LLM Logs/
    ├── ChatGPT Logs/    # Imported history
    └── Claude Code Logs/ # Raw transcripts

Scripts

chatgpt-import.py

Imports ChatGPT conversation exports into Obsidian.

# Export from ChatGPT: Settings → Data Controls → Export Data
unzip ~/Downloads/chatgpt-export.zip conversations.json -d /tmp/
python3 scripts/chatgpt-import.py

raw_log.js

Converts Claude Code JSONL transcripts to readable markdown.

# Manual export (add as alias for convenience)
TRANSCRIPT=$(ls -t ~/.claude/projects/-home-*/*.jsonl | grep -v agent | head -1)
SESSION_ID=$(basename "$TRANSCRIPT" .jsonl)
echo "{\"session_id\":\"$SESSION_ID\",\"transcript_path\":\"$TRANSCRIPT\"}" | node ~/.claude/hooks/raw_log.js

Requirements

• Claude Code CLI
• Obsidian (or any markdown-based notes system)
• Python 3.6+
• Node.js

Known Limitations

• No auto-export: Raw transcript export is manual (hook doesn't auto-fire on session end)
• Summary logs depend on Claude: If Claude forgets or ignores CLAUDE.md, no logs
• Not a plugin: This is a workflow system using plain files, not an Obsidian plugin

License

MIT - do whatever you want with these.

Contributing

Issues and PRs welcome. This is a personal workflow shared publicly - your mileage may vary.