Turn your notes into interactive AI conversations. Chat with multiple AI providers, split topics into linked cards, and explore your knowledge graph visually — all without leaving Obsidian.
Latest: v0.4.0 — Now with virtual scrolling, conversation export, and keyboard shortcuts!
Chat with your favorite AI models — all in one place:
| Provider | Models |
|---|---|
| Anthropic | Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku |
| OpenAI | GPT-4o, GPT-4 Turbo, GPT-3.5 Turbo |
| Google Gemini | Gemini 1.5 Pro, Gemini 1.5 Flash |
| Qwen | Qwen 2.5 72B, Qwen 2.5 32B |
| DeepSeek | DeepSeek Chat V3 |
| Kimi | Kimi Chat |
✅ Per-provider API keys — Configure each provider independently and switch freely without re-entering credentials.
- Every conversation is a Markdown file — fully editable, version-controllable, and portable
- Split topics — Branch any assistant message into a linked child card with one click
- Link related ideas — Connect cards manually; links appear as wikilinks in Graph View
- Visualize connections — Watch your knowledge graph grow in Obsidian's Graph View
- Auto-suggest context — AI automatically recommends relevant cards from your vault
- Manual context picker — Curate which notes inform each conversation
- Smart context injection — Your selected cards are injected as conversation context
- Token-by-token responses — Watch AI responses appear in real time
- Interrupt anytime — Stop generation mid-response if you have what you need
- 📥 Export Conversations — Export as PDF, Markdown, or plain text
- ⌨️ Keyboard Shortcuts — Ctrl+L (focus input), Ctrl+, (settings), Ctrl+K (search)
- 🚀 Virtual Scrolling — Smooth scrolling even with 1000+ messages
- ⚡ Performance — 60% faster load times, debounced context suggestions
⏳ Pending approval — once approved, install directly from Obsidian:
- Open Settings → Community Plugins
- Click Browse
- Search for "Neural Chat"
- Click Install → Enable
- Install the BRAT plugin from Community Plugins
- Open BRAT settings → Add a beta plugin
- Enter repo:
howlil/obsidian-neural-chat - Click Add Plugin → Enable after installation
- Download
main.js,manifest.json, andstyles.cssfrom the latest release - Copy all files into
<YourVault>/.obsidian/plugins/obsidian-neural-chat/ - Reload Obsidian → Settings → Community Plugins → Enable "Neural Chat"
- Open Settings → Neural Chat
- Select your AI Provider from the dropdown (Anthropic, OpenAI, Gemini, etc.)
- Enter your API Key for the selected provider
- Choose a Model (defaults to provider's recommended model)
- Cards folder: Where chat files are stored (default:
neural-cards/) - Files are plain Markdown with YAML frontmatter — fully portable
| Setting | Description | Default |
|---|---|---|
| Max context cards | Maximum cards AI can suggest for context | 5 |
| Max context messages | Maximum historical messages sent to AI | 20 |
| System prompt | Custom system prompt for all conversations | (built-in) |
- Press
Ctrl/Cmd + Pto open Command Palette - Run: Neural Chat: New Chat Card
- A new card opens — type your message and press Enter to send
💡 Tip: Cards are stored in your configured folder with auto-generated IDs and topic-based titles.
- Hover over any assistant message — a Split button appears
- Select text you want to branch (optional — uses full message if nothing selected)
- Click Split → AI generates a topic name automatically
- A child card is created with:
- Link back to parent card
- Selected content as starting context
- Opens in a new tab
- Click the link icon 🔗 in the card header
- Search cards by topic or filename
- Click a card to link it
- Link appears as
[[wikilink]]in both cards' Markdown — visible in Graph View
- Context bar (below header) shows active context cards
- Click + to open Context Picker and add cards manually
- Click × on a badge to remove it
- On load, AI auto-suggests relevant cards if context is empty
Each card is a plain Markdown file with YAML frontmatter:
YourVault/
└── neural-cards/
├── 20260315120000-learning-rust.md ← "Learning Rust"
├── 20260315120500-ownership-borrowing.md ← child of above
└── 20260315121000-generics-lifetimes.md ← linked sibling
---
id: 20260315120000-learning-rust
topic: Learning Rust
createdAt: 2026-03-15T12:00:00Z
parentCard: null
childCards: []
linkedCards: []
contextCards: []
---
## Conversation
**You:** How do I get started with Rust?
**Assistant:** Great question! Rust is a systems programming language...✅ Fully editable — Edit cards directly in Obsidian; changes sync automatically.
- Your data stays yours — Conversations are stored locally in your vault
- API keys never leave your device — Keys stored in Obsidian's encrypted config
- No telemetry — No analytics, no tracking, no external calls except to your chosen AI provider
- Bring your own API key — You pay only for your own usage directly to the provider
Q: Which provider should I use?
A: Depends on your needs:
- Anthropic Claude — Best for nuanced reasoning and long context
- OpenAI GPT-4 — Great all-rounder with strong coding skills
- Google Gemini — Excellent for multimodal tasks (future update)
- Qwen/DeepSeek/Kimi — Cost-effective alternatives with strong performance
Q: Is my conversation data sent anywhere besides the AI provider?
A: No. Messages go only to your selected provider's API using your own key. No third-party servers.
Q: What happens if my API quota runs out?
A: The plugin shows an error banner. All existing cards remain safe on disk — no data loss.
Q: Can I edit card files directly in Obsidian?
A: Yes! The plugin re-reads files on every vault modify event, so edits sync automatically.
Q: Can I switch providers mid-conversation?
A: Yes — switch in Settings, but note that conversation history is provider-agnostic text, so it works with any provider.
# Clone the repo
git clone https://github.com/howlil/obsidian-neural-chat.git
cd obsidian-neural-chat
# Install dependencies
pnpm install
# Build
pnpm build
# Run tests
pnpm test
# Lint
pnpm lint- Language: TypeScript + React
- Bundler: esbuild
- Tests: Jest
- Package Manager: pnpm
Contributions welcome! See CONTRIBUTING.md for guidelines.
- 🐛 Report bugs via Issues
- 💡 Request features via Discussions
- 🔧 Submit PRs against the
masterbranch
MIT License — Use freely for personal or commercial projects.
- Built with Obsidian — A second brain, for you, forever.
- AI providers: Anthropic, OpenAI, Google, Alibaba Qwen, DeepSeek, Kimi
Made with ❤️ by Mhd Ulil Abshar