Vault Chat Agent

AI assistant plugin for working with your Obsidian vault.

What Works

• One agentic chat with Ask and Edit intents.
• OpenAI-compatible chat completions via /v1/chat/completions.
• Search scopes for whole-vault, current-note, and current-folder chat.
• Plan mode for Edit chats, allowing read-only investigation before pending edits are prepared.
• @current, @[[Note]], @path.md, and @folder/ context mentions with inline autocomplete.
• Provider presets for OpenAI, DeepSeek, OpenRouter, LM Studio, Ollama, and custom endpoints.
• Model browsing from OpenAI-compatible /v1/models endpoints.
• Local vault indexing across supported text files, Canvas files, and metadata-only attachments.
• Graph-aware retrieval using note links, backlinks, and neighboring chunks.
• Related Notes sidebar based on the active note or selected text.
• Read-only agent tools for searching, opening notes, inspecting links, and reviewing the current note, with clickable source references.
• Text commands for summarizing selections, extracting action items, and proposing polished rewrites.
• Saved prompts with Ask/Edit intent, runnable from the command palette against the current selection or note.
• Manual workflows for weekly reviews, meeting tasks, and project status notes, all behind pending edit review.
• Reviewed edit proposals with patch previews, Apply/Reject, and batch patch support.
• Optional stripping of <think>, <reasoning>, and <thought> blocks from assistant output.
• Realtime re-indexing on vault create, modify, delete, and rename events.

Build

npm install
npm run build

Then copy the contents of dist/ into:

<vault>/.obsidian/plugins/vault-chat-agent/

Enable the plugin in Obsidian community plugin settings, then configure an OpenAI-compatible provider in the plugin settings.

Provider Notes

The plugin calls:

<API base URL>/v1/chat/completions

Remote providers usually need an API key. Local providers such as LM Studio and Ollama can leave the API key empty when using a localhost base URL.