CluEdit

Desktop application for browsing, searching, backing up, branching, and exporting Claude Code conversations as production-ready LLM training data. Built with Tauri + SvelteKit.

Features

Conversation Browser

• Browse all Claude Code projects and conversations from ~/.claude/projects/
• Rich metadata: message counts, token usage, cost estimates, tools used, files modified
• Full-text search powered by Tantivy
• Filter by date range, sort by modified/created/size
• Live refresh via filesystem watcher — new conversations appear automatically

Conversation Viewer

• Formatted message display with syntax-highlighted code blocks (Shiki)
• Expandable tool call details with input/output
• In-conversation search
• Stats dashboard with token counts and estimated costs
• Continuation chain navigation between related conversations

Backup & Branching

• Backup conversations at any point — full or truncated at a specific event
• Restore from any backup with automatic safety backup of current state
• Branch conversations — duplicate with all UUIDs regenerated so Claude treats the branch as independent
• Branch from any message — click "branch here" on any message to fork from that point
• Branch from backups to create new conversations from saved checkpoints

Training Data Export

Export conversations in formats ready for LLM fine-tuning:

Format	Mode	Description
ChatML	Conversational	OpenAI fine-tuning JSONL — text-only exchanges, no tool noise
ChatML + Tools	Agentic	Structured tool_calls + role: "tool" for training tool-calling models
ShareGPT	Conversational	human/gpt turn pairs for open-source fine-tuning frameworks
Alpaca	Conversational	Instruction/output pairs for instruction-tuning

Training exports include:

• Automatic stripping of Claude-specific XML tags and system artifacts
• Filtering of low-value narration messages
• Merging consecutive same-role messages for valid alternating format
• Token-aware chunking — long conversations split into segments under 16k tokens
• Proper OpenAI function-calling schema with tools array for agentic mode

Also exports to JSON, Markdown, and plain text.

Export Scope

• Single conversation — from the conversation viewer
• All conversations in a project — from the toolbar
• All projects — from the sidebar

Getting Started

Prerequisites

• Rust (stable)
• Node.js 18+
• pnpm

Setup

pnpm install
pnpm tauri dev

Build

pnpm tauri build

Project Structure

src/                          # SvelteKit frontend
  lib/
    components/               # Svelte 5 components
    stores/                   # Svelte stores
    api.ts                    # Tauri IPC bindings
    types.ts                  # TypeScript types
src-tauri/                    # Rust backend
  src/
    backup_service.rs         # Backup, restore, branch with ID remapping
    content_sanitizer.rs      # Training data cleaning, chunking, tool schemas
    conversation_service.rs   # Core I/O, metadata, export (8 formats)
    conversation_analyzer.rs  # Metadata extraction from JSONL events
    search_indexer.rs         # Tantivy full-text search index
    file_watcher.rs           # Live filesystem monitoring
    commands.rs               # Tauri IPC command handlers
    models.rs                 # Shared data types

Tech Stack

• Framework: Tauri 2.x
• Backend: Rust (serde, tantivy, regex, tokio, uuid)
• Frontend: SvelteKit 2.x, Svelte 5, TypeScript
• Styling: Tailwind CSS 4, Bits UI, Lucide icons
• Code Highlighting: Shiki

License

MIT