AI Chat Extract

A CLI tool to extract, organize, and index chat exports from AI platforms (ChatGPT, Claude, etc.) into browsable, searchable folders.

The Problem

AI chat exports are unwieldy:

• Large JSON blobs (e.g., 736 conversations, 78MB)
• Too big for direct context loading
• Nested structures that are hard to navigate
• No organization by date, project, or topic

The Solution

Stream-process exports into organized markdown files with full asset extraction:

output/
├── index.json                    # Quick lookup manifest
├── projects/
│   └── my-project/
│       ├── 2024-12-10-chat-about-feature/
│       │   ├── conversation.md
│       │   └── assets/
│       └── ...
└── conversations/
    ├── 2024-07/
    │   ├── 2024-07-14-financial-planning/
    │   │   ├── conversation.md
    │   │   ├── artifacts/
    │   │   └── assets/
    │   └── ...
    └── 2024-12/
        └── ...

Features

• Stream processing - Memory-efficient parsing of large exports
• Multi-platform - Parsers for ChatGPT, Claude, and more
• Project-aware - Chats belonging to projects stay organized together
• Full extraction - Conversations, artifacts, attachments, memories
• Searchable index - JSON manifest for quick lookups

Tech Stack (Phase 1 - MVP)

Component	Choice	Purpose
CLI	Typer + Rich	Beautiful command-line interface
Parsing	ijson	Memory-efficient JSON streaming
Models	Pydantic	Data validation and structure
Output	Markdown	Universal, portable format
Container	Docker	Consistent deployment

Future Phases

• Phase 2: PostgreSQL for cross-conversation queries / Qdrant vector DB for semantic search ("find similar chats")

Usage

# Extract ChatGPT export
ai-extract chatgpt ./input/chatgpt-export.zip -o ./output

# Extract Claude export
ai-extract claude ./input/claude-export.json -o ./output

# Generate index only (no file extraction)
ai-extract index ./input/export.json

Supported Platforms

Platform	Export Format	Status
ChatGPT	ZIP (conversations.json + assets)	🔲 Planned
Claude	JSON	🔲 Planned
Claude Projects	JSON	🔲 Planned

AI Instructions

We use spec-kit here. See the .specify folder for workflow details.

Development

# Install dependencies
uv sync

# Run CLI
uv run ai-extract --help

# Run with Docker
docker compose up