Mirror your AI. Own your knowledge.
Git for AI conversations. AI conversations are valuable, but today they disappear into proprietary chat histories. Obsidian LLMHub captures your AI conversations, converts them into structured Markdown, enriches them with metadata, links them into your knowledge graph, and stores everything in your own repository — searchable, version-controlled, and future-proof.
Claude · ChatGPT · Gemini · Codex · Perplexity · Cursor · Windsurf · Open WebUI
│
Conversation Normalizer
│
Knowledge Pipeline
Markdown · YAML · Chunk · Tag · Link · (Summary · Embedding)
│
Obsidian · Logseq · Git · Filesystem · (Notion)
See docs/DESIGN.md for the full architecture and roadmap.
pip install -e .
# 1. Create a vault (any directory; git init it for version history)
obsidian-llmhub init ~/ai-vault
git -C ~/ai-vault init
# 2. Import your AI conversation exports (format auto-detected)
obsidian-llmhub import ~/Downloads/claude-export/conversations.json --vault ~/ai-vault
obsidian-llmhub import ~/Downloads/chatgpt-export/conversations.json --vault ~/ai-vault
# 3. Inspect
obsidian-llmhub status --vault ~/ai-vaultRe-running an import is idempotent: unchanged conversations are skipped, updated ones are rewritten, and each sync becomes one clean Git commit.
conversations/<source>/<YYYY>/<MM>/<slug>-<id>.md— one Markdown note per conversation with YAML frontmatter (id, source, dates, tags), role-labeled sections, and[[wiki-links]]to related conversations.index.md— an auto-generated map of content, newest first..obsidianllmhub/chunks/*.jsonl— embedding-ready chunks for RAG tooling.
Open the vault in Obsidian or Logseq and your AI history becomes a browsable, linked second brain.
| Source | Status |
|---|---|
| Claude.ai data export | ✅ |
| ChatGPT data export | ✅ |
| Generic JSON (canonical schema — bring any source) | ✅ |
| Gemini / Perplexity / Cursor / Windsurf / Open WebUI | 🚧 roadmap |
pip install -e . pytest
pytestMIT