Hey Supernote community! I've been building a plugin that lets you lasso-select anything on your note and have a full multi-turn conversation with an AI about it. Sharing it here in case others find it useful.

What it does

Select any content on your note page with the lasso tool, tap the SuperAI button in the toolbar, choose how the AI should read it (image or text), and start chatting. The AI remembers the full conversation so you can ask follow-ups naturally.

Supported AI providers

You bring your own API key — the plugin works with:

• Google Gemini (free tier available at aistudio.google.com)
• OpenAI (GPT-4o, o4-mini, etc.)
• Anthropic Claude (Claude Opus, Sonnet, Haiku)
• OpenRouter (one key, access to hundreds of models from many providers)

First-time setup

1. Install the .snplg file via the Supernote file manager
2. Open any note and tap the SuperAI button in the regular toolbar (the gear/settings button)
3. Choose your AI provider
4. Paste your API key using the 📋 Paste button
5. Tap ↓ Load Models to fetch the latest available models from that provider, then tap the one you want
6. Tap Save & Start Chat

The key and model are saved — you won't need to do this again unless you want to switch providers.

How to use it

For Image AI:

1. Lasso-select any content on your page (handwriting, diagrams, equations, sketches — anything)
2. Tap the SuperAI lasso toolbar button that appears
3. Choose Image AI — the selection is captured as a visual snapshot and sent to the AI
4. The AI responds with an analysis; type follow-up questions in the chat box
5. Use the Copy button on any message, or 📋 in the header to copy the whole conversation

For Text AI:

1. Same lasso selection step
2. Choose Text AI — the Supernote device runs its own handwriting recognition on the selection, and that recognized text is sent to the AI instead of an image
3. Faster, cheaper (smaller API payload), and often more accurate for dense written text

Use case ideas

Image AI — best for visual content

• Lasso a hand-drawn diagram or flowchart and ask "explain what this process shows" or "what's missing from this flow?"
• Select a sketch or mind map and ask "turn this into a structured outline"
• Capture a page of mixed notes + diagrams and ask "summarize the key points from this page"
• Select a table you drew by hand and ask "what patterns do you see in these numbers?"
• Lasso a hand-drawn graph or chart and ask "what trend is this showing?"
• Select a page of equations or physics diagrams and ask "walk me through this step by step"
• Grab a rough wireframe sketch and ask "describe this UI layout and suggest improvements"
• Circle a section of your Zettelkasten note and ask "how does this connect to concept X?"

Text AI — best for dense written content

• Select a page of meeting notes and ask "give me 3 action items from this"
• Lasso a block of writing and ask "is this argument logically coherent?"
• Select your study notes on a topic and ask "quiz me on this" then keep chatting through the quiz
• Grab a paragraph and ask "rewrite this more concisely" or "translate this to Japanese"
• Select a to-do list you wrote and ask "prioritize these by urgency"
• Lasso a page of brainstorming and ask "which of these ideas are most novel?"
• Select a problem statement you wrote and ask "what are three approaches to solving this?"
• Grab your notes on a book/paper and ask "what questions am I still not answering?"
• Select a decision you're weighing (pros/cons written out) and ask "what's the strongest argument on each side?"

Multi-turn conversations (both modes)

• Start with a summary, then drill in: "tell me more about point 2"
• Get a first draft, then iterate: "make it shorter" / "make it more formal"
• Ask it to generate something, then refine: "now add examples"

A few notes

• Each new lasso selection starts a completely fresh conversation
• The settings button (⚙) is accessible from both the chat header and the error screen if you want to switch providers mid-session
• Image AI works even if no text is recognized (diagrams, pure sketches, etc.)
• Text AI will be greyed out and disabled if Supernote's recognizer can't find any text in the selection — in that case just use Image AI

Happy to answer questions or take feedback. Still iterating on it!

Download Link
https://drive.google.com/file/d/1zzlptzI6rvlUvHabmG4EcoVPVuBHJeih/view?usp=sharing