Chatting with AI

Chatting with AI

An agentic AI assistant that lives in your Obsidian vault — same experience on phone, tablet, and desktop.

🌐 English · 简体中文 · 繁體中文 · 日本語

---

✨ Why?

• Three providers, your choice — Anthropic API, OpenAI API, or sign in with your ChatGPT account. No marketplace of half-broken providers.
• 14 vault-native tools — read, edit, search, create, rename, frontmatter, backlinks. The agent goes from idea to changed files without leaving the chat.
• Mobile-parity, by design — no streaming, no Node-only modules, no localhost callbacks. iOS and Android behave the same as desktop.
• Selection scope — highlight text in any note, send it to chat, the assistant edits only inside the selection.
• Secrets in the OS keychain — never in data.json, never synced across devices.

🎬 One conversation, many tools

Ask once, the agent figures out which tools to call:

You: Find every note in /Books that's missing a `rating` property and add `rating: ?`.

Assistant
  → search_vault("/Books")               → 12 files
  → get_properties("Books/Sapiens.md")   → has rating
  → get_properties("Books/Hail Mary.md") → no rating
  → set_properties("Books/Hail Mary.md", { rating: "?" })
  → ... (5 more)

  Done — added `rating: ?` to 6 notes:
  - Books/Hail Mary.md
  - Books/Klara and the Sun.md
  - ...

The agent reads before it edits, prefers small surgical changes over rewrites, and doesn't re-ask once you confirm.

🎯 Selection scope

Highlight text in any note, right-click, choose Send selection to Chat. The selection becomes a pill above the input and the assistant edits only inside the selected range — leaving the rest of the document byte-identical.

[ pill: "...the introduction was a bit dry, and..."  ✕ ]

You: tighten this — keep my voice

The agent uses find-and-replace scoped to the selection text. Everything outside the highlight stays untouched.

🛠️ 14 vault-native tools

⚙️ Providers

Attach up to four images to a message, including by pasting an image from the clipboard. Images are sent with that message to the selected provider; large photos are resized before sending.

About ChatGPT account login. This provider signs you in with your ChatGPT account and routes requests through the ChatGPT/Codex backend (not api.openai.com). It requires an active ChatGPT plan with Codex access. The model picker fetches the Codex catalog for your account; account permissions still determine whether a model can be used.

Model list updates

All three providers load their authenticated model catalogs in settings. Lists are cached for 24 hours and survive restarts. Settings show the last list immediately, then refresh it in the background when expired. No daily timer or request on every settings visit is scheduled. The refresh button checks immediately. Failed requests retain the last list and selected model; account/API-key changes use a separate cache.

ChatGPT OAuth discovers the latest stable client version from OpenAI's official GitHub release metadata and keeps the last known version or bundled fallback if discovery fails. Reasoning effort, reasoning summaries and parallel-tool options follow catalog metadata where provided. New model IDs normally require no plugin update, but upstream protocol changes can still require one. Custom model IDs survive reload. Catalog data contains no credentials.

🚀 Quick start

1. Open Settings → Community plugins → Browse.
2. Search for Chatting with AI.
3. Click Install, then Enable.
4. Settings → Chatting with AI → pick a provider, paste an API key (or click Connect ChatGPT).
5. Open the chat from the ribbon icon or the command palette.

📦 Install

Community Plugins (recommended)

This is the default install path now that Chatting with AI is in the official Obsidian Community Plugins directory.

1. In Obsidian, open Settings → Community plugins.
2. If prompted, turn off Restricted Mode / turn on community plugins.
3. Click Browse and search for Chatting with AI.
4. Click Install, then Enable.

Migrating from BRAT

If you previously installed the beta through BRAT, move to the community version once it appears in the plugin browser:

1. In Settings → Community plugins → Installed plugins, disable Chatting with AI.
2. Open BRAT settings and remove o1xhack/obsidian-chatting from the beta plugin list.
3. Return to Settings → Community plugins → Browse and search for Chatting with AI.
4. If Obsidian shows it as Installed, open its entry and click Enable. If it shows Install, click Install, then Enable.
5. Open Settings → Chatting with AI and confirm your provider settings are still present.

The plugin migrates data from the old obsidian-chatting folder to chatting-with-ai on first load. If you used a very old beta and the community browser does not recognize it as installed, installing the community version is still safe; the migration runs when the new plugin starts.

1. Download main.js, manifest.json, styles.css from the latest release
2. Place them in <vault>/.obsidian/plugins/chatting-with-ai/
3. Reload Obsidian and enable Chatting with AI in Community Plugins

git clone https://github.com/o1xhack/obsidian-chatting.git
cd obsidian-chatting
npm install
npm run build

Symlink into a test vault:

ln -s "$(pwd)" /path/to/vault/.obsidian/plugins/chatting-with-ai

🧭 Design principles

🗺️ Roadmap

• Three providers (Anthropic, OpenAI, ChatGPT account)
• 14 vault-native tools
• iOS / Android parity
• Selection scope
• Official Obsidian Community Plugins listing
• Multi-conversation history with archive / search
• Image attachments where the provider supports them
• Dynamic provider catalogs with automatic refresh when the 24-hour cache expires

Have a request? Open an issue.

❓ FAQ

Only what the agent needs for the current turn. When you ask a question, the agent decides which tools to call — read_document, search_vault, etc. — and the contents fetched by those calls (plus the active note context) are sent to your chosen provider. Images you attach are sent with that message and retained in local chat history until you clear the conversation. Nothing is uploaded in the background. There is no vault index.

Yes — that's the design constraint everything else is built around. Requests go through Obsidian's requestUrl() (mobile WebViews enforce CORS), there's no streaming, no Node-only modules, no localhost callback for OAuth. iOS and Android run the same code path as desktop.

It uses your existing ChatGPT plan (Plus, Pro, Team, Enterprise) — there's no separate billing. You need an active plan with Codex access. The plugin doesn't talk to api.openai.com; it talks to the same backend the Codex CLI uses.

Probably not — keeping the provider list small is a deliberate choice. Two API providers cover the major ecosystems, and ChatGPT account login covers the "I just have a ChatGPT plan" case. Adding more would mean more permutations to validate on mobile.

Locally in <vault>/.obsidian/plugins/chatting-with-ai/chat-state.json. It is not synced by Obsidian Sync (plugin data files are excluded by default). API keys live in the OS keychain via SecretStorage and are also not synced. Attached image data is retained in this local chat history until you clear the conversation.

🤝 Contributing

Issues and PRs welcome. Before opening a PR:

• Run npx tsc --noEmit and npm run svelte-check
• Test on at least one mobile platform (iOS or Android) — the mobile-parity rule is real
• For larger changes, open an issue first to align on direction

🙏 Acknowledgements

Originally derived from omarshahine/obsidian-chat (MIT). The original copyright is preserved in LICENSE. Chatting with AI is now an independent project with its own roadmap — major rewrites include the agent loop, mobile-parity work, the ChatGPT account provider, and the selection-scope feature.

📄 License

MIT.

---

Author: Yuxiao (o1xhack) · app.o1xhack.com