Obsidian Conversation Catalog

Catalog all Claude conversations (browser exports + Claude Code CLI sessions) into an Obsidian vault with rich frontmatter, Claude-generated summaries, and Dataview-compatible structure.

Quick Start

Sync All Conversations

python ~/.claude/skills/obsidian-catalog/scripts/sync.py --all

Import Browser Export

python ~/.claude/skills/obsidian-catalog/scripts/sync.py --browser "/path/to/export.zip"

Sync Claude Code History Only

python ~/.claude/skills/obsidian-catalog/scripts/sync.py --code

Filter by Date

python ~/.claude/skills/obsidian-catalog/scripts/sync.py --code --since 2025-12-01

List Sessions Without Syncing

python ~/.claude/skills/obsidian-catalog/scripts/sync.py --all --list-only

Configuration

Edit ~/.claude/skills/obsidian-catalog/config.yaml to configure:

• vault_path: Path to Obsidian vault's Claude folder
• claude_code_path: Path to Claude Code history (default: ~/.claude/projects)
• auto_summarize: Whether to generate summaries automatically

Generating Summaries

After syncing, generate summaries for sessions that need them:

1. Run sync to export sessions (without summaries)
2. Ask: "Generate summaries for the synced sessions"
3. I will use sub agents to process sessions in batches
4. Summaries are written to each session's frontmatter

For large imports (100+ sessions), summaries are generated in batches using sub agents to manage context efficiently.

Searching Conversation History

Once synced, find past sessions by asking:

• "What did we work on last Tuesday?"
• "Find sessions about GIS processing"
• "Show recent conversations about the land dashboard"
• "What topics have we discussed this month?"

I will search the Obsidian vault at the configured path and return relevant sessions.

Output Structure

Sessions are organized in the vault as:

Claude/
├── Sessions/
│   ├── Browser/2025/01/24/     # Claude.ai conversations
│   │   └── 2025-01-24-960f89e7.md
│   └── Code/2025/12/06/        # Claude Code sessions
│       └── 2025-12-06-2d31c8e7.md
├── Summaries/
│   ├── Daily/                   # Daily rollups
│   └── Projects/                # Per-project summaries
├── _raw/                        # Original export data
└── _Dashboard.md                # Dataview queries

Frontmatter Fields

Each session file includes:

• type: claude-session
• source: browser | code
• session_id: UUID
• title: Conversation name
• date, created, updated: Timestamps
• project: Project name
• message_count: Number of messages
• topics: Keywords (Claude-generated)
• summary: Brief summary (Claude-generated)
• has_code: Whether it contains code blocks
• word_count, duration_minutes: Metrics

Sub Agent Strategy

For summary generation, I use sub agents (Task tool) to:

• Process sessions in batches of 10-20
• Maintain fresh context for each batch
• Generate summaries in parallel when possible
• Handle 300+ conversations without context overflow