NanoGPT Assistant for Obsidian

A full-featured Obsidian plugin integrating NanoGPT's AI capabilities including chat, text generation, image generation, web search, web scraping, and note editing.

Features

• AI Chat Assistant: Streaming chat with per-note history stored in plugin settings
• Ask/Edit Modes: Ask questions or edit the active note directly with inline accept/reject
• Safe Edit Mode: Apply targeted edits without rewriting the full note
• Edit Backups: Automatic timestamped backups before applying AI edits
• Inline Autocomplete: Ghost-text suggestions with Tab-to-accept
• Folder Context: Optionally include other notes in the same folder
• Text Generation: Summarize, expand, rewrite, and explain text
• Image Generation: Create images using NanoGPT image models
• Web Search: Search the web and insert formatted results with citations
• Web Scraping: Paste any URL in chat to scrape and save pages as notes
• Note Creation: Create and link multiple notes from chat edit mode
• Model Filtering: Filter models by subscription tier
• Image Storage: Automatically save generated images to vault

Installation

1. Download the latest release from GitHub Releases
2. Extract the archive to your Obsidian vault's plugins folder: YourVault/.obsidian/plugins/obsidian-nanogpt/
3. Enable the plugin in Obsidian Settings > Community Plugins
4. Click on "NanoGPT Assistant" in the settings to configure your API key

Setup

1. Get your NanoGPT API key from nano-gpt.com/api
2. Open Obsidian Settings > Community Plugins > NanoGPT Assistant
3. Enter your API key in the settings
4. Configure default model, temperature, and other options as needed

Usage

Chat

1. Open any note
2. Click the NanoGPT ribbon icon or run "NanoGPT: Open Chat" command
3. Choose Ask or Edit mode in the header
4. Type your message and click Send
5. Edit mode returns a proposed update with Accept/Reject buttons
6. Chat history is stored in plugin settings per note path

Edit mode commands

• create note: Title - Create a new note in the current folder
• create notes: A, B, C - Create multiple notes and link them back

Text Generation Commands

• NanoGPT: Generate Summary - Summarize selected text
• NanoGPT: Expand Selection - Expand on selected content
• NanoGPT: Rewrite Text - Rewrite selected text
• NanoGPT: Explain Selection - Explain in simpler terms
• NanoGPT: Inline Autocomplete - On-demand completion at cursor

Image Generation

• NanoGPT: Generate Image - Open modal to generate image
• NanoGPT: Generate Image from Selection - Use selected text as prompt
• Generated images can be saved to vault or inserted directly into notes

Web Search

• NanoGPT: Web Search - Search web and insert results
• NanoGPT: Web Search from Selection - Use selected text as query
• Results include formatted markdown with citations

Web Scraping

• Paste a URL into chat to scrape and save it as a note
• A confirmation prompt appears before scraping
• Scraped notes are linked back into the active note

Note Completion

• NanoGPT: Complete Note - Generate content based on note title
• NanoGPT: Generate Outline - Create outline for note

Configuration

API Settings

• API Key: Your NanoGPT API key (required)
• Base URL: NanoGPT API endpoint (default: https://nano-gpt.com/api/v1)
• Default Model: Model to use for chat/text (default: zai-org/glm-4.7-flash)
• Show Subscription Only: Filter to subscription-included models

Chat Settings

• System Prompt: Default system prompt for chat
• Temperature: Controls randomness (0.0 - 2.0, default: 0.7)
• Max Tokens: Maximum tokens to generate (default: 2048)
• Safe Edit Mode: Use targeted edits instead of rewriting the full note
• Backup Notes Before AI Edits: Save a timestamped copy before applying AI edits
• Edit Backup Folder: Vault folder to store backups (default: .nanogpt-backups)

Image Settings

• Image Size: Default size for generated images (256x256 to 1536x1024)
• Image Quality: Standard or HD quality

Web Search Settings

• Web Search Results: Number of results to return (default: 5)
• Web Search Provider: Provider name for NanoGPT web search
• Web Search Options: JSON options sent to the provider
• Include Citations: Add citation section to results

Autocomplete Settings

• Inline Autocomplete: Enable ghost-text suggestions
• Autocomplete Delay: Wait time before suggestions

Folder Context Settings

• Folder Context: Include other notes in the active folder
• Folder Context Max Notes: Limit number of notes included
• Folder Context Max Characters: Total context cap

Chat History Storage

Chat history is stored in the plugin's saved settings, keyed by note path. This keeps note files untouched while preserving per-note context.

Image Storage

Generated images are saved to your vault's attachments folder (respects your Obsidian attachment settings). Images are named with format: nanogpt-{timestamp}-{hash}.png

Default Model

The default model is zai-org/glm-4.7-flash. You can change this in settings or use any other NanoGPT-supported model.

Development

# Install dependencies
npm install

# Development build with hot reload
npm run dev

# Production build
npm run build