No description
Find a file
2025-12-04 23:10:16 -08:00
.github/workflows ci: remove broken workflow 2025-12-04 23:10:16 -08:00
src fix: get token usage from OpenAI API 2025-11-23 20:31:01 -08:00
.editorconfig Initial commit 2025-01-07 15:24:13 -08:00
.eslintignore Initial commit 2025-01-07 15:24:13 -08:00
.eslintrc Initial commit 2025-01-07 15:24:13 -08:00
.gitignore chore: ignore aider files 2025-03-04 15:28:25 -08:00
.npmrc Initial commit 2025-01-07 15:24:13 -08:00
demo.gif docs: update readme 2025-01-15 21:46:28 -08:00
esbuild.config.mjs feat: launch 1st version 2025-01-15 21:18:52 -08:00
LICENSE Create LICENSE 2025-01-15 22:01:54 -08:00
manifest.json build: bump version to 1.5.0 2025-11-21 09:05:56 -08:00
package-lock.json feat: add support for custom API providers with OpenAI SDK 2025-10-21 21:03:02 -07:00
package.json feat: add support for custom API providers with OpenAI SDK 2025-10-21 21:03:02 -07:00
README.md docs: update readme 2025-11-21 09:04:20 -08:00
styles.css Initial commit 2025-01-07 15:24:13 -08:00
tsconfig.json Initial commit 2025-01-07 15:24:13 -08:00
version-bump.mjs Initial commit 2025-01-07 15:24:13 -08:00
versions.json build: bump version to 1.5.0 2025-11-21 09:05:56 -08:00

Claude AI for Obsidian

CAO allows you to chat with Claude directly in a note.

Features

  1. Chat in notes with editable content
  2. Manage chat histories as plain notes
  3. Customize chat options in front matter
  4. Use wikilinks for notes as chat context
  5. Support custom API providers such as OpenRouter
  6. Support custom prompts with templates
  7. Support image embedding in prompts

Installation

Search for CAO in Obsidian's community plugins page.

Usage

  • Set up API providers
    • You can choose to use official Anthropic APIs or OpenAI-compatible API providers such as OpenRouter
    • For official Anthropic APIs, enter your API key and select a model from the dropdown menu
    • For custom API providers, you need to also set the base URL(it defaults to https://api.openai.com/v1 for OpenAI models)
      • In this case, you need to manually enter the model name of which the format may vary depending on the provider
  • Config chat options
    • You can designate a custom folder for chat histories
    • You can also update max tokens, temperature and system prompt to suit your needs
    • Toggle streaming response open for better performance and showing stats allows you to see token usage for each response
  • Try available commands
    • Use Open new chat command to create a new chat note
    • Optionally, there's Add/Reset chat options for customization
    • For replies, fire up Get response(you may want a hot key for this, such as Cmd/Ctrl + .)
    • Next time, use Open last chat to resume last conversation or Select chat to select a history chat

Here're the available chat options with example values to set in the front matter:

---
model: claude-sonnet-4-5
max_tokens: 1024
temperature: 1
system_prompt: You are a helpful AI assistant
---

Custom API providers

CAO supports OpenAI-compatible API providers. Here's an example for OpenRouter:

  • Base URL: https://openrouter.ai/api/v1
  • Model format: anthropic/claude-sonnet-4.5 or openai/gpt-4o
  • Get your API key from OpenRouter

Notes:

  • Using Claude models with custom API providers may have limitations.
  • It's possible to use non-Claude models now, but be aware it is experimental and not fully tested.

Custom prompts

Now you can add custom prompts in settings, which will become available in the command palette or as slash commands(if toggled on in core plugins).

Additionally, you can use {cursor} in the prompt template as a placeholder of cursor for editing after the prompt is inserted.

Supported Models

CAO supports all Claude models, take a look at here for more details.

You can either choose an active model in the dropdown menu, or specify an old one in the front matter as illustrated above.

Contributing

For bug fixes or feature improvements, please create an issue or pull request.

For ideas or any other questions, please post in discussion.

Any suggestions or support is welcome and appreciated, thank you!