Sidekick

AI assistant that lives inside your Obsidian vault

Chat with 33 tools · 8 agent presets · inline autocomplete · 3 providers

Report Bug · Request Feature

---

Sidekick is a multi-provider AI chat assistant for Obsidian, primarily inspired by the VS Code GitHub Copilot Chat extension. It brings the same powerful chat + tools + inline autocomplete workflow to your Obsidian vault. It opens as a sidebar panel and can search your vault, read and create notes, generate images, browse the web, and more — all through natural conversation. It works with OpenAI, OpenRouter, and GitHub Copilot (free tier included).

Quick Start

1. Install the plugin (instructions below)
2. Add your API key in Settings → Sidekick
3. Click the chat icon in the sidebar — done

Why Sidekick?

Most AI plugins give you a chatbox and nothing else. Sidekick gives the AI 33 tools to actually interact with your vault — search, read, create, edit, move, and delete notes; browse files and tags; fetch URLs; generate images; and even search Reddit and Jira. Combined with agent presets and inline autocomplete, it's less of a chat plugin and more of a co-pilot for your knowledge base.

Your data stays yours. Sidekick sends messages only to the provider you choose. No telemetry, no analytics, no third-party servers. Conversations are saved as plain markdown files in your vault.

Features

Chat & Conversation

• Streaming responses with real-time thinking/reasoning display
• Chat persistence — conversations saved as markdown in your vault
• Conversation management — multiple chats, pinning, search, export
• Custom slash commands — define your own /commands with custom system prompts

Tools & Agents

• 33 built-in tools — the AI decides what to use based on your request
• 8 agent presets — switch between specialized personas in one click
• Sub-agents — delegate complex multi-step tasks
• Tool approval — risky tools (create/delete/fetch) require your confirmation

Intelligence

• User profiling — learns your preferences over time for better responses
• Auto-RAG — automatically finds relevant notes via embeddings before answering
• Smart search — BM25 + fuzzy matching + graph boost + recency weighting

Writing

• Inline autocomplete — ghost text suggestions as you type (Tab to accept, Escape to dismiss)
• Diff view — see proposed edits before applying them

Providers

• OpenAI — GPT-4o, GPT-4.1, o3, o4, DALL-E, and more
• OpenRouter — 200+ models from Anthropic, Google, Meta, Mistral, DeepSeek, etc.
• GitHub Copilot — works with your existing Copilot subscription (including free tier)

Screenshots

Provider & Model Settings

Tool Calling

Inline Auto-Completion

Embeddings & Vector Search

Agent Presets

Switch personas instantly for different tasks:

Preset	Description
🤖 Default	General-purpose assistant
💻 Code Expert	Senior engineer — reviews, optimizes, explains code
✍️ Writing Coach	Improves clarity, tone, and structure
🔬 Research Assistant	Deep-dives into topics with citations
🎓 Socratic Tutor	Teaches through questions, not answers
💡 Brainstorm Partner	Generates and riffs on ideas
📝 Markdown Editor	Formats, restructures, and polishes notes
🔧 Debugger	Finds and fixes bugs step by step

Tools

The AI can call these tools autonomously during conversation:

Category	Tools
Vault	search_vault · read_note · read_note_section · read_note_outline · create_note · edit_note · move_note · delete_note · open_note
Browse	list_files · grep_search · search_by_tag · get_recent_notes · get_open_notes · get_backlinks · get_note_metadata
Web	web_search · fetch_url · search_reddit · read_reddit_post · jira_search · jira_get_issue · jira_create_issue · jira_add_comment · jira_update_issue
Media	generate_image · view_image
AI	semantic_search_vault · delegate_to_agent · spawn_parallel_agents
User	ask_user · ask_user_choice · remember_user_fact

Slash Commands

Command	Description
/help	Show command list
/note	Attach active note as context
/selection	Attach selected text as context
/regen	Regenerate last response
/iterate	Toggle iterate mode
/clear	Clear current chat
/export	Export chat to a markdown note
/new	Start a new conversation
/rename	Rename current conversation
/duplicate	Duplicate current conversation
/model	Open model picker
/settings	Open plugin settings
/usage	Show Copilot usage quota
/pin	Toggle pin on current conversation
/info	Show current conversation info
/stats	Show vault-wide chat statistics
/favorites	Show favorited (thumbs-up) messages
/search	Search across all conversations
/undo	Remove last user message and response
/summary	Show conversation summary (messages, cost, model)
/agent	Switch agent preset
/profile	Show learned user profile

Installation

BRAT (Recommended for now)

1. Install BRAT from Community Plugins
2. BRAT → Add Beta Plugin → anwar3606/obsidian-sidekick-ai
3. Enable "Sidekick" in Community Plugins

Manual

1. Download main.js, manifest.json, and styles.css from the latest release
2. Create <your-vault>/.obsidian/plugins/sidekick/
3. Copy the three files into that folder
4. Restart Obsidian → Settings → Community plugins → Enable "Sidekick"

Configuration

OpenAI

1. Get an API key from platform.openai.com
2. Settings → Sidekick → Provider: OpenAI → paste your key

OpenRouter

1. Get an API key from openrouter.ai
2. Settings → Sidekick → Provider: OpenRouter → paste your key
3. Access to 200+ models including Claude, Gemini, Llama, Mistral, and more

GitHub Copilot

1. Settings → Sidekick → Provider: GitHub Copilot
2. Click "Sign in with GitHub" — authenticates via OAuth device flow
3. Works with any Copilot plan including the free tier

Privacy

• No telemetry. Sidekick collects nothing.
• No middleman. Messages go directly from your machine to the provider you chose (OpenAI, OpenRouter, or GitHub).
• Local storage. Conversations are saved as markdown files in your vault — visible, searchable, and yours.
• Open source. Every line of code is auditable in this repository.

FAQ

Does it work on mobile?

Yes. The plugin is not desktop-only. However, inline autocomplete (ghost text) is currently desktop-only due to CodeMirror limitations on mobile.

Which provider should I use?

• GitHub Copilot — easiest to start with (free tier, no credit card). Limited to models GitHub offers.
• OpenAI — best for GPT models and DALL-E image generation. Pay-as-you-go.
• OpenRouter — most model variety (200+). Pay-as-you-go with very competitive pricing.

Can I use local/self-hosted models?

If your local server exposes an OpenAI-compatible API (e.g., Ollama, LM Studio, vLLM), you can point OpenAI or OpenRouter provider URLs to it.

Where are conversations stored?

As markdown files in your vault under Sidekick/conversations/ (configurable in settings). They're regular notes — you can search, link, and edit them.

Contributing

Contributions are welcome! The codebase is split into two layers:

• lib/ — Pure logic with zero Obsidian dependencies (independently testable)
• src/ — Obsidian UI layer that imports from lib/

# Requires Node.js >= 25 and pnpm
pnpm install
pnpm dev          # watch mode
pnpm build        # production build
pnpm test         # 1,500+ tests

License

MIT