InsightPage - Features

InsightPage Assistant: Key Features

InsightPage integrates an AI chat engine, an interactive floating assistant, a local markdown vault, and smart web clipping into a unified research and knowledge management workflow.

1. AI Chat Engine

InsightPage’s side panel features a flexible, powerful chat interface that connects directly to local or cloud-based AI models.

Chat Modes

  • Standard Chat Mode (chat): Converse with the AI using general knowledge and conversation history.
  • Web Mode (web): The AI optimizes your query, performs a live web search, visits top relevant pages to extract content, and synthesizes a fully-sourced, up-to-date answer.
  • Page Mode (page): The AI reads the active browser tab’s content (including web PDFs) to summarize, extract structured data, or answer questions in context.

Type /r before a query (e.g. /r key takeaways from project alpha) to run a Hybrid Search across your notes:

  • Keyword Search (BM25): Matches exact keywords in notes and chats.
  • Semantic Search (Embeddings): Matches the underlying conceptual meaning.
  • MMR Reranking: Re-ranks results to maximize relevance while minimizing redundancy before feeding context to the LLM.

Smart Dispatcher

When given multi-step requests, the smart_dispatcher acts as your personal AI agent:

  1. Searches the web.
  2. Extracts full article text.
  3. Synthesizes structured summaries.
  4. Saves findings directly to your Note Viewer vault.

2. Floating Assistant

The Floating Assistant brings AI assistance directly onto any webpage, letting you interact with content seamlessly without opening the full side panel.

graph LR
    FAB[Floating Action Button] -->|Click| FW[Floating Window]
    FW -->|Summary| QA[Automatic Page Summary & Q&A]
    FW -->|Drag / Resize| Custom[Saved Layout Preferences]
    FW -->|Export| SidePanel[Transfer Chat to Side Panel]

    style FAB fill:#e1f5fe,stroke:#0288d1;
    style FW fill:#e1f5fe,stroke:#0288d1;
  • Floating Action Button (FAB): A lightweight button floating at the edge of the viewport. Drag vertically to reposition anytime.
  • Floating Window: Movable, resizable overlay window.
  • In-Context Actions:
    • Automatic Summaries: Instant page overview, detected tags, and suggested follow-up questions.
    • In-Context Chat: Ask questions directly about the page content.
    • Quick Transfers: Clear chat with Refresh, or migrate the conversation to the main side panel with Open in Side Panel.

3. Note Viewer & Personal Vault

The note-taking system serves as your second brain, offering full Markdown compatibility and native local-file integration.

graph TD
    NV[Note System] -->|Double Click / Open| Tab[Full-Tab Note Editor]
    Tab -->|AI Edit Request| Diff[Diff View: Original vs. AI suggested]
    Diff -->|Accept / Decline| Final[.md File Saved Directly to PC]

    style NV fill:#fff3e0,stroke:#f57c00;
    style Tab fill:#fff3e0,stroke:#f57c00;
  • Local Markdown Vault: Connect folders on your PC directly using the File System Access API. Notes are saved as standard .md files in real-time, functioning as a lightweight Obsidian-like setup. No intermediate cache layer is used, preventing write amplification and keeping your local directory as the single source of truth.
  • Obsidian Style & Rich Visualization Ecosystem:
    • InsightPage’s rendering capabilities are powered by the official Obsidian community plugin 3-ark/insightpage-style. Use the plugin in Obsidian to get the exact same charts, maps, timelines, ABC sheet music audio player, and inline popover comments in your desktop Obsidian vault.
  • Wikilinks & Deep Internal Linking ([[Note Title]]):
    • Full Wikilink syntax support across notes, popovers, and chat responses.
    • Type [[Note Title]] to link to any note in your vault. Clicking a Wikilink navigates directly or creates the note if it does not exist yet.
  • Note Mentions (@[Note Title]) & Context Injection:
    • Type @ in the chat input to search and attach specific notes directly into the AI context window.
  • Select Notes for Q&A:
    • Select multiple notes from your Vault list and click Start Q&A to pre-fill note mentions (@[Note Title]) in the chat bar. The AI will answer queries using strictly the selected notes.
  • Full-Page Tab Editor: Open any note in a new browser tab for a spacious, distraction-free editing layout complete with adjustable font sizes.
  • AI-Powered Note Editing & Diff View:
    1. Open your note in a new tab.
    2. In the side panel, ask the AI to modify the note (e.g., “Make this summary more concise”, “Translate to German”).
    3. View the real-time Diff View in your note tab, highlighting additions and deletions side-by-side.
    4. Click Accept to commit the changes directly to your file, or Decline to keep the original.

Explore all visual rendering features in our Rendering Showcase.

4. Context Window & Smart Clipping

  • Context Window Integration: Drag files, images, or enter specific URLs into the chat’s input context window. URLs are automatically scraped and used as additional prompt context.
  • Smart Web Clipper:
    • Highlight any text on a webpage, right-click, and select Add to Note to append the text to your current popover note.
    • Use the context menus to quickly capture insights without breaking your browsing momentum.
  • Smart Weekly Notes:
    • Instantly create or open weekly review files with predefined action templates.
    • Highlight text on any page, right-click, and select Add to Weekly Note to instantly append clean, sanitized web links (stripping analytical and tracking parameters) alongside your captured snippets.
    • Customize action items and path folders under Settings -> Page Mode Settings.

5. Advanced & Multi-Modal Capabilities

Multimodal Image & Vision Model Support

  • Vision-Ready Uploads: Upload standard image formats (.png, .jpg, .jpeg, .webp, .gif) via the chat input interface.
  • Automatic Resolution & Quality Optimization: Large images (longer side > 1568px) are automatically scaled down while maintaining aspect ratio. Large PNG images are automatically optimized to JPEG (90% quality with white background fallbacks) before dispatch to avoid API limits.
  • Native Vision Payloads: Images are sent as structured, OpenAI-compatible image_url vision payloads, compatible with cloud endpoints (GPT-4o) and local multi-modal hosts (LM Studio).

Intelligent PDF Operations & Language Auto-Detection

  • Dynamic Language Detection: Automatically detects language families—including English, Spanish, German, French, Simplified Chinese (zh-CN), Traditional Chinese (zh-TW), Japanese, and Korean.
  • One-Click PDF Generation: The note action menu features a direct PDF download button that automatically selects appropriate system fonts and runs high-performance regex checks for CJK rendering based on text content.
  • Background PDF Text Extraction: Web PDF tab text parsing is handled asynchronously in background service workers and offscreen documents, preventing UI freezes and bundle bloat.

Theme-Aware Design & Floating Window Polish

  • Unified Custom Theme Palette Mapping: When using themes like Moss, Night Sepia, or Dark, all dropdown selectors, settings panels, hover menus, and modals adapt seamlessly.
  • Theme-Aware Floating Tooltips: Tooltips leverage CSS variables (--tooltip-bg, --tooltip-fg, and --tooltip-border) to match the active persona theme.
  • Hover-Triggered Action Overlays: Quick-action controls (such as web search toggles) transition smoothly on container hover or input focus.

6. Privacy, Security & Data Ownership

  • 100% Local-First Architecture: No developer servers, no proxies, no user registration, and no tracking.
  • Direct Client-to-API Communication: Browser talks directly to your configured AI provider or local server.
  • Learn more on our Privacy & Security Policy.